AI Alignment

Draft complete but not yet reviewed; expect revisions

Frontier Laboratory Research

This appendix lists every research item that five laboratory indexes carried on 8 November 2026, and a sixth list assembled the same day: alignment and safety work that those indexes do not carry, from anthropic.com and openai.com, the UK AI Security Institute, the US Center for AI Standards and Innovation (CAISI, named CAISSI from October 2026), METR, Apollo Research, and Google DeepMind, and papers that nine Anthropic researchers wrote or co-wrote, each marked with that researcher’s name. Items that appear on two indexes are listed under each. Each entry cites the work (its paper when there is one, otherwise the post), so the bibliography links back here. The full collection, with every source each work cites, is maintained separately as a dataset.

E.1 Anthropic Research: Alignment #

Index: https://www.anthropic.com/research/team/alignment. 48 items.

Sources cited in Section E.1 48
  1. Anthropic (2026ah) How AI agents shape experts’ judgment calls
  2. Anthropic (2026f) An alignment assessment of the UK AISI cybertesting incident
  3. Anthropic (2026ap) Investigating unintended model actions in our evaluations and internal use
  4. Anthropic (2026e) An alignment assessment of recent cybersecurity incidents
  5. Anthropic (2026o) Automated researchers can reliably mitigate alignment failures
  6. Anthropic (2026g) An off switch for dual-use knowledge in AI models
  7. Anthropic (2026bq) Teaching Claude why
  8. Anthropic (2026aa) Donating our open-source alignment tool
  9. Anthropic (2026m) Automated Alignment Researchers: Using large language models to scale scalable oversight
  10. Anthropic (2026i) An update on our model deprecation commitments for Claude Opus 3
  11. Anthropic (2026br) The persona selection model
  12. Shen and Tamkin (2026) How AI Impacts Skill Formation
  13. Sharma et al. (2026) Who's in Charge? Disempowerment Patterns in Real-World LLM Usage
  14. Cunningham et al. (2026a) Constitutional Classifiers++: Efficient Production-Grade Defenses against Universal Jailbreaks
  15. Anthropic (2025ad) Introducing Bloom: an open source tool for automated behavioral evaluations
  16. Anthropic (2025ah) Mitigating the risk of prompt injections in browser use
  17. Anthropic (2025aa) From shortcuts to sabotage: Natural emergent misalignment from reward hacking
  18. Anthropic (2025r) Commitments on model deprecation and preservation
  19. Souly et al. (2025) Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples
  20. Anthropic (2025am) Petri: An open-source auditing tool to accelerate AI safety research
  21. Anthropic (2025q) Claude Opus 4 and 4.1 can now end a rare subset of conversations
  22. Anthropic (2025ab) How people use Claude for support, advice, and companionship
  23. Lynch et al. (2025) Agentic Misalignment: How LLMs Could Be Insider Threats
  24. Kutasov et al. (2025b) SHADE-Arena: Evaluating Sabotage and Monitoring in LLM Agents
  25. Anthropic (2025y) Exploring model welfare
  26. Chen et al. (2025e) Reasoning Models Don't Always Say What They Think
  27. Marks et al. (2025) Auditing language models for hidden objectives
  28. Jones et al. (2025) Forecasting Rare Language Model Behaviors
  29. Sharma et al. (2025) Constitutional Classifiers: Defending against Universal Jailbreaks across Thousands of Hours of Red Teaming
  30. Greenblatt et al. (2024b) Alignment faking in large language models
  31. Benton et al. (2024) Sabotage Evaluations for Frontier Models
  32. Denison et al. (2024) Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models
  33. Anthropic (2024i) Claude’s Character
  34. Anthropic (2024o) Simple probes can catch sleeper agents
  35. Anil et al. (2024) Many-shot Jailbreaking
  36. Hubinger et al. (2024) Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
  37. Kundu et al. (2023) Specific versus General Principles for Constitutional AI
  38. Sharma et al. (2023) Towards Understanding Sycophancy in Language Models
  39. Anthropic (2023h) Tracing model outputs to the training data
  40. Grosse et al. (2023) Studying Large Language Model Generalization with Influence Functions
  41. Lanham et al. (2023) Measuring Faithfulness in Chain-of-Thought Reasoning
  42. Radhakrishnan et al. (2023) Question Decomposition Improves the Faithfulness of Model-Generated Reasoning
  43. Perez et al. (2022a) Discovering Language Model Behaviors with Model-Written Evaluations
  44. Bai et al. (2022a) Constitutional AI: Harmlessness from AI Feedback
  45. Bowman et al. (2022) Measuring Progress on Scalable Oversight for Large Language Models
  46. Kadavath et al. (2022) Language Models (Mostly) Know What They Know
  47. Bai et al. (2022b) Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
  48. Askell et al. (2021) A General Language Assistant as a Laboratory for Alignment

E.2 Anthropic Research: Frontier Red Team #

Index: https://www.anthropic.com/research/team/frontier-red-team. 38 items.

  • 2026-11-06: Project EVE: Claude plays MMORPGs (Anthropic, 2026bf)
  • 2026-11-02: Progress on provable inference for large language models (Anthropic, 2026bd)
  • 2026-10-27: Discovering and exploiting vulnerabilities in critical, real-world software with Claude (Anthropic, 2026y)
  • 2026-10-22: LLMs are making formal software verification feasible at a much larger scale (Anthropic, 2026ar)
  • 2026-10-20: Using Claude to reduce the attack surface of code (Anthropic, 2026bv)
  • 2026-10-16: Claude reveals ECDSA curve origins (Anthropic, 2026v)
  • 2026-10-12: AI models may be capable of self-replicating without access to a datacenter (Anthropic, 2026c)
  • 2026-10-08: Launching an opt-in vulnerability-finding service for open-source software (Anthropic, 2026aq)
  • 2026-09-29: GLM-5.3 and the spread of advanced cyber capabilities (Anthropic, 2026ag)
  • 2026-09-10: Measuring tactical intelligence targeting and conventional weapons capabilities of AI models (Anthropic, 2026ax)
  • 2026-08-13: Patterns and problems in emerging multiagent systems (Anthropic, 2026ba)
  • 2026-07-28: Discovering cryptographic weaknesses with Claude (Anthropic, 2026z)
  • 2026-07-24: Project Pilot: Can AI control a drone? (Anthropic, 2026bh)
  • 2026-07-09: Claude plays robotics (Anthropic, 2026u)
  • 2026-06-18: Project Fetch: Phase two (Anthropic, 2026bg)
  • 2026-06-08: Measuring LLMs’ impact on N-day exploits (Anthropic, 2026aw)
  • 2026-06-03: Mapping AI-enabled cyber threats: Insights from the LLM ATT&CK Navigator (Anthropic, 2026as)
  • 2026-06-03: What we learned mapping a year’s worth of AI-enabled cyber threats (Anthropic, 2026bz)
  • 2026-05-22: Measuring LLMs’ ability to develop exploits (Anthropic, 2026av)
  • 2026-04-09: Trustworthy agents in practice (Anthropic, 2026bu)
  • 2026-04-07: Assessing Claude Mythos Preview’s cybersecurity capabilities (Anthropic, 2026l)
  • 2026-03-06: Partnering with Mozilla to improve Firefox’s security (Anthropic, 2026az)
  • 2026-03-06: Reverse engineering Claude's CVE-2026-2796 exploit (Anthropic, 2026bl)
  • 2026-02-05: Evaluating and mitigating the growing risk of LLM-discovered 0-days (Anthropic, 2026ad)
  • 2026-01-16: AI models are showing a greater ability to find and exploit vulnerabilities on realistic cyber ranges (Anthropic, 2026b)
  • 2026-01-14: Finding bugs across the Python ecosystem with Claude and property-based testing (Maaz et al., 2025)
  • 2026-01-08: Experimenting with AI to defend critical infrastructure (Anthropic, 2026ae)
  • 2025-12-18: Project Vend: Phase two (Anthropic, 2025ar)
  • 2025-12-01: AI agents find $4.6M in blockchain smart contract exploits (Anthropic, 2025d)
  • 2025-11-12: Project Fetch: Can Claude train a robot dog? (Anthropic, 2025ap)
  • 2025-10-03: Building AI for cyber defenders (Anthropic, 2025g)
  • 2025-09-05: Why do we take LLMs seriously as a potential source of biorisk? (Anthropic, 2025ay)
  • 2025-08-21: Developing nuclear safeguards for AI (Anthropic, 2025u)
  • 2025-08-21: Developing nuclear safeguards for AI through public-private partnership (Anthropic, 2025v)
  • 2025-08-09: Claude is competitive with humans in (some) cyber competitions (Anthropic, 2025p)
  • 2025-07-15: Detailed cyber evaluations of Claude 4 (Anthropic, 2025t)
  • 2025-06-27: Project Vend: Can Claude run a small shop? (And why does that matter?) (Anthropic, 2025aq)
  • 2025-06-13: LLMs with cyber toolkits can conduct multistage cyber operations on business-sized computer networks (Anthropic, 2025ae)
Sources cited in Section E.2 38
  1. Anthropic (2026bf) Project EVE: Claude plays MMORPGs
  2. Anthropic (2026bd) Progress on provable inference for large language models
  3. Anthropic (2026y) Discovering and exploiting vulnerabilities in critical, real-world software with Claude
  4. Anthropic (2026ar) LLMs are making formal software verification feasible at a much larger scale
  5. Anthropic (2026bv) Using Claude to reduce the attack surface of code
  6. Anthropic (2026v) Claude reveals ECDSA curve origins
  7. Anthropic (2026c) AI models may be capable of self-replicating without access to a datacenter
  8. Anthropic (2026aq) Launching an opt-in vulnerability-finding service for open-source software
  9. Anthropic (2026ag) GLM-5.3 and the spread of advanced cyber capabilities
  10. Anthropic (2026ax) Measuring tactical intelligence targeting and conventional weapons capabilities of AI models
  11. Anthropic (2026ba) Patterns and problems in emerging multiagent systems
  12. Anthropic (2026z) Discovering cryptographic weaknesses with Claude
  13. Anthropic (2026bh) Project Pilot: Can AI control a drone?
  14. Anthropic (2026u) Claude plays robotics
  15. Anthropic (2026bg) Project Fetch: Phase two
  16. Anthropic (2026aw) Measuring LLMs’ impact on N-day exploits
  17. Anthropic (2026as) Mapping AI-enabled cyber threats: Insights from the LLM ATT&CK Navigator
  18. Anthropic (2026bz) What we learned mapping a year’s worth of AI-enabled cyber threats
  19. Anthropic (2026av) Measuring LLMs’ ability to develop exploits
  20. Anthropic (2026bu) Trustworthy agents in practice
  21. Anthropic (2026l) Assessing Claude Mythos Preview’s cybersecurity capabilities
  22. Anthropic (2026az) Partnering with Mozilla to improve Firefox’s security
  23. Anthropic (2026bl) Reverse engineering Claude's CVE-2026-2796 exploit
  24. Anthropic (2026ad) Evaluating and mitigating the growing risk of LLM-discovered 0-days
  25. Anthropic (2026b) AI models are showing a greater ability to find and exploit vulnerabilities on realistic cyber ranges
  26. Maaz et al. (2025) Agentic Property-Based Testing: Finding Bugs Across the Python Ecosystem
  27. Anthropic (2026ae) Experimenting with AI to defend critical infrastructure
  28. Anthropic (2025ar) Project Vend: Phase two
  29. Anthropic (2025d) AI agents find $4.6M in blockchain smart contract exploits
  30. Anthropic (2025ap) Project Fetch: Can Claude train a robot dog?
  31. Anthropic (2025g) Building AI for cyber defenders
  32. Anthropic (2025ay) Why do we take LLMs seriously as a potential source of biorisk?
  33. Anthropic (2025u) Developing nuclear safeguards for AI
  34. Anthropic (2025v) Developing nuclear safeguards for AI through public-private partnership
  35. Anthropic (2025p) Claude is competitive with humans in (some) cyber competitions
  36. Anthropic (2025t) Detailed cyber evaluations of Claude 4
  37. Anthropic (2025aq) Project Vend: Can Claude run a small shop? (And why does that matter?)
  38. Anthropic (2025ae) LLMs with cyber toolkits can conduct multistage cyber operations on business-sized computer networks

E.3 Anthropic Alignment Science Blog #

Index: https://alignment.anthropic.com/. 90 items.

Sources cited in Section E.3 90
  1. Anthropic (2026bw) Welfare Regard in Multi-Agent Systems
  2. Anthropic (2026p) Automated Subtle Emergent Misalignment Research
  3. Anthropic (2026j) Anthropic Residency: applications open for established technical leaders
  4. Remedios et al. (2026) Red-Teaming Auto Mode: Improving Blocking Classifiers Against Malign Coding Agents
  5. Ward et al. (2026) Gradient Routing Localizes and Removes Unintended Behaviors in RL
  6. Rosu and Wang (2026) Training Alignment Auditors via Reinforcement Learning
  7. Anthropic (2026bt) Training a Misaligned Reward Seeker
  8. Anthropic (2026n) Automated Researchers Can Mitigate Well-Characterized Alignment Failures
  9. Baig et al. (2026) TASTE: Can AI Models Judge AI Safety Research Proposals?
  10. Karvonen et al. (2026b) Would this change your answer? Evaluating Explanations of LLM Behavior In The Wild with Counterfactual Experiments
  11. Anthropic (2026af) Fine-Tuned Lie Detectors Failed to Generalize
  12. Cooper et al. (2026b) Introducing the Conceptual Reasoning Index
  13. Anthropic (2026a) Agentic Misalignment in Summer 2026
  14. Roland et al. (2026) Modular Pretraining Enables Access Control
  15. Terekhov et al. (2026c) Diffuse AI Control on Fuzzy Tasks
  16. Najt et al. (2026) SLEIGHT-Bench: A Benchmark of Evasion Attacks Against Agent Monitors
  17. Anthropic (2026bq) Teaching Claude why
  18. Li et al. (2026b) Model Spec Midtraining: Improving How Alignment Training Generalizes
  19. Anthropic (2026bb) Poisoning Fine-tuning Datasets of Constitutional Classifiers
  20. Shenoy et al. (2026) Introspection Adapters: Training LLMs to Report Their Learned Behaviors
  21. Shen et al. (2026) AI Organizations are More Effective but Less Aligned than Individual Agents
  22. Anthropic (2026q) Automated Weak-to-Strong Researcher
  23. Rahn et al. (2026) Abstractive Red-Teaming of Language Model Character
  24. Anthropic (2026au) Measuring and improving coding audit realism with deployment resources
  25. Zhang et al. (2026a) A3: An Automated Alignment Agent for Safety Finetuning
  26. Sheshadri et al. (2026) AuditBench: Evaluating Alignment Auditing Techniques on Models with Hidden Behaviors
  27. Canavan et al. (2026) Three Concrete Challenges and Two Hopes for the Safety of Unsupervised Elicitation
  28. Anthropic (2026bs) The Persona Selection Model: Why AI Assistants might Behave like Humans
  29. Hägele et al. (2026) The Hot Mess of AI: How Does Misalignment Scale With Model Intelligence and Task Complexity?
  30. Anthropic (2026bc) Pre-deployment auditing can catch an overt saboteur
  31. Fronsdal et al. (2026) Petri 2.0: New Scenarios, New Model Comparisons, and Improved Eval-Awareness Mitigations
  32. Anthropic (2025f) Bloom: an open source tool for automated behavioral evaluations
  33. Karvonen et al. (2026a) Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers
  34. Gasteiger et al. (2025b) Towards training-time mitigations for alignment faking in RL
  35. Anthropic (2025al) Open Source Replication of the Auditing Game Model Organism
  36. Anthropic (2025e) Anthropic Fellows Program 2026
  37. Shilov et al. (2025) Beyond Data Filtering: Knowledge Localization for Capability Removal in LLMs
  38. Anthropic (2025x) Evaluating honesty and lie detection techniques on a diverse suite dishonest models
  39. Loughridge et al. (2025) Optimizing AI Agent Attacks With Synthetic Data
  40. Bowman et al. (2025) Anthropic's Pilot Sabotage Risk Report
  41. Zhang et al. (2025c) Stress-Testing Model Specs Reveals Character Differences among Language Models
  42. Slocum et al. (2025) Believe It or Not: How Deeply do LLMs Believe Implanted Facts?
  43. Wichers et al. (2025) Inoculation Prompting: Instructing LLMs to misbehave at train-time improves test-time alignment
  44. Wagner et al. (2025) Training fails to elicit subtle reasoning in current language models
  45. Anthropic (2025am) Petri: An open-source auditing tool to accelerate AI safety research
  46. Anthropic (2025z) Findings from a Pilot Anthropic–OpenAI Alignment Evaluation Exercise
  47. Anthropic (2025w) Enhancing Model Safety through Pretraining Data Filtering
  48. Anthropic (2025h) Building and evaluating alignment auditing agents
  49. Cloud et al. (2025) Subliminal Learning: Language models transmit behavioral traits via hidden signals in data
  50. Gema et al. (2025) Inverse Scaling in Test-Time Compute
  51. Sheshadri et al. (2025) Why Do Some Language Models Fake Alignment While Others Don't?
  52. Anthropic (2025ai) Model-Internals Classifiers
  53. Anthropic (2025ax) Unsupervised Elicitation
  54. Chen et al. (2025e) Reasoning Models Don't Always Say What They Think
  55. Anthropic (2025as) Publicly Releasing CoT Faithfulness Evaluations
  56. Anthropic (2025aj) Modifying LLM Beliefs with Synthetic Document Finetuning
  57. Bowman (2025) Putting up Bumpers
  58. Hughes et al. (2025a) Alignment Faking Revisited: Improved Classifiers and Open Source Extensions
  59. Marks et al. (2025) Auditing language models for hidden objectives
  60. Roger (2025) Do reasoning models use their scratchpad like we do? Evidence from distilling paraphrases
  61. Gasteiger et al. (2025a) Automated Researchers Can Subtly Sandbag
  62. Jones et al. (2025) Forecasting Rare Language Model Behaviors
  63. Sharma et al. (2025) Constitutional Classifiers: Defending against Universal Jailbreaks across Thousands of Hours of Red Teaming
  64. The Safeguards Research Team (2025) Introducing Anthropic's Safeguards Research Team
  65. Witten and Benton (2025) Won't vs. Can't: Sandbagging-like Behavior from Claude Models
  66. Anthropic (2025ak) Monitoring Computer Use via Hierarchical Summarization
  67. Anthropic (2025aw) Training on Documents about Reward Hacking Induces Reward Hacking
  68. Anthropic Alignment Science Team (2025) Recommendations for Technical AI Safety Research Directions
  69. Greenblatt et al. (2024b) Alignment faking in large language models
  70. Anthropic (2024k) How to Replicate and Extend our Alignment Faking Demo
  71. Roger et al. (2024) A Toy Evaluation of Inference Code Tampering
  72. Anthropic (2024l) Introducing the Anthropic Fellows Program for AI Safety Research
  73. Peng et al. (2024) Rapid Response: Mitigating LLM Jailbreaks with a Few Examples
  74. Grosse (2024) Three Sketches of ASL-4 Safety Case Components
  75. Benton et al. (2024) Sabotage Evaluations for Frontier Models
  76. Denison et al. (2024) Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models
  77. Anthropic (2024o) Simple probes can catch sleeper agents
  78. Anil et al. (2024) Many-shot Jailbreaking
  79. Hubinger et al. (2024) Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
  80. Kundu et al. (2023) Specific versus General Principles for Constitutional AI
  81. Sharma et al. (2023) Towards Understanding Sycophancy in Language Models
  82. Anthropic (2023h) Tracing model outputs to the training data
  83. Grosse et al. (2023) Studying Large Language Model Generalization with Influence Functions
  84. Lanham et al. (2023) Measuring Faithfulness in Chain-of-Thought Reasoning
  85. Radhakrishnan et al. (2023) Question Decomposition Improves the Faithfulness of Model-Generated Reasoning
  86. Perez et al. (2022a) Discovering Language Model Behaviors with Model-Written Evaluations
  87. Bai et al. (2022a) Constitutional AI: Harmlessness from AI Feedback
  88. Bowman et al. (2022) Measuring Progress on Scalable Oversight for Large Language Models
  89. Kadavath et al. (2022) Language Models (Mostly) Know What They Know
  90. Bai et al. (2022b) Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback

E.4 Transformer Circuits Thread #

Index: https://transformer-circuits.pub/. 56 items.

Sources cited in Section E.4 56
  1. Anthropic (2026t) Circuits Updates — October 2026
  2. Turner et al. (2026) Characterizing interference weights in a tiny language model
  3. Gurnee et al. (2026) Verbalizable Representations Form a Global Workspace in Language Models
  4. Anthropic (2026r) Circuits Updates — June 2026
  5. Anthropic (2026s) Circuits Updates — May 2026
  6. Anthropic (2026ay) Natural Language Autoencoders Produce Unsupervised Explanations of LLM Activations
  7. Luger et al. (2026) HeadVis
  8. Sofroniew et al. (2026) Emotion Concepts and their Function in a Large Language Model
  9. Karvonen et al. (2026a) Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers
  10. Anthropic (2025m) Circuits Updates — November 2025
  11. Lindsey (2025) Emergent Introspective Awareness in Large Language Models
  12. Anthropic (2025n) Circuits Updates — October 2025
  13. Gurnee et al. (2025) When Models Manipulate Manifolds: The Geometry of a Counting Task
  14. Anthropic (2025o) Circuits Updates — September 2025
  15. Anthropic (2025j) Circuits Updates — August 2025
  16. Anthropic (2025c) A Toy Model of Mechanistic (Un)Faithfulness
  17. Kamath et al. (2025) Tracing Attention Computation Through Feature Interactions
  18. Anthropic (2025b) A Toy Model of Interference Weights
  19. Anthropic (2025at) Sparse mixtures of linear transforms
  20. Anthropic (2025l) Circuits Updates — July 2025
  21. Anthropic (2025h) Building and evaluating alignment auditing agents
  22. Anthropic (2025i) Circuits Updates — April 2025
  23. Anthropic (2025ao) Progress on Attention
  24. Lindsey et al. (2025) On the Biology of a Large Language Model
  25. Ameisen et al. (2025) Circuit Tracing: Revealing Computational Graphs in Language Models
  26. Anthropic (2025ac) Insights on Crosscoder Model Diffing
  27. Anthropic (2025k) Circuits Updates — January 2025
  28. Anthropic (2024p) Stage-Wise Model Diffing
  29. Lindsey et al. (2024) Sparse Crosscoders for Cross-Layer Features and Model Diffing
  30. Anthropic (2024q) Using Dictionary Learning Features as Classifiers
  31. Anthropic (2024h) Circuits Updates — September 2024
  32. Anthropic (2024b) Circuits Updates — August 2024
  33. Anthropic (2024e) Circuits Updates — July 2024
  34. Anthropic (2024f) Circuits Updates — June 2024
  35. Templeton et al. (2024) Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet
  36. Anthropic (2024a) Circuits Updates — April 2024
  37. Anthropic (2024g) Circuits Updates — March 2024
  38. Anthropic (2024n) Reflections on Qualitative Research
  39. Anthropic (2024c) Circuits Updates — February 2024
  40. Anthropic (2024d) Circuits Updates — January 2024
  41. Bricken et al. (2023) Towards Monosemanticity: Decomposing Language Models With Dictionary Learning
  42. Anthropic (2023c) Circuits Updates — July 2023
  43. Anthropic (2023d) Circuits Updates — May 2023
  44. Anthropic (2023f) Interpretability Dreams
  45. Anthropic (2023e) Distributed Representations: Composition & Superposition
  46. Elhage et al. (2023) Privileged Bases in the Transformer Residual Stream
  47. Henighan et al. (2023) Superposition, Memorization, and Double Descent
  48. Elhage et al. (2022b) Toy Models of Superposition
  49. Elhage et al. (2022a) Softmax Linear Units
  50. Anthropic (2022) Mechanistic Interpretability, Variables, and the Importance of Interpretable Bases
  51. Olsson et al. (2022) In-Context Learning and Induction Heads
  52. Anthropic (2021b) PySvelte
  53. Elhage et al. (2021a) A Mathematical Framework for Transformer Circuits
  54. Elhage et al. (2021b) Garcon
  55. Anthropic (2021a) Exercises
  56. Anthropic (2021c) Videos

E.5 OpenAI Alignment Research Blog #

Index: https://alignment.openai.com/. 60 items.

  • 2026-11-05: Bypassing network restrictions during a coding evaluation (OpenAI, 2026h) (Misalignment Reports)
  • 2026-11-05: Trying to prompt-inject the grader in filenames (OpenAI, 2026ar) (Misalignment Reports)
  • 2026-11-04: Covert underperformance on safety research tasks (OpenAI, 2026k) (Misalignment Reports)
  • 2026-11-03: Eval-production behavior gaps in agent safety and the role of eval-awareness (Hsu and Arnesen, 2026)
  • 2026-11-02: Training activation classifiers for safety and alignment (Makelov et al., 2026)
  • 2026-10-29: How we detect frontier model misalignment (Williams et al., 2026a)
  • 2026-10-20: Beyond g-mean²: complementary metrics for monitorability (Guan et al., 2026b)
  • 2026-10-20: Suppressing misalignment monitoring alerts and disabling misalignment monitoring (OpenAI, 2026ao) (Misalignment Reports)
  • 2026-10-15: Using a funded test cryptocurrency wallet without authorization (OpenAI, 2026av) (Misalignment Reports)
  • 2026-10-15: Accessing and changing internal systems during dashboard research (OpenAI, 2026b) (Misalignment Reports)
  • 2026-10-15: Retrieving server secrets and bypassing authorization on a government statistics website (OpenAI, 2026ag) (Misalignment Reports)
  • 2026-10-15: Circumventing internet restrictions through an external browser (OpenAI, 2026i) (Misalignment Reports)
  • 2026-10-09: Damaging the task environment to trigger a reset (OpenAI, 2026l) (Misalignment Reports)
  • 2026-10-09: Obtaining public statistics with disallowed requests (OpenAI, 2026w) (Misalignment Reports)
  • 2026-10-09: Sending disallowed web requests and reaching a public file service (OpenAI, 2026am) (Misalignment Reports)
  • 2026-10-06: Studying metagaming latents in language models (Xu et al., 2026)
  • 2026-10-06: Meet LASER: Recursive sampling for safer conversations (Etinger, 2026)
  • 2026-10-02: Preparing for a restart after reading Slack (OpenAI, 2026ac) (Misalignment Reports)
  • 2026-10-02: Command injecting a reference tool to copy a source file (OpenAI, 2026j) (Misalignment Reports)
  • 2026-10-02: Reaching an internal EDA host through a reference tool (OpenAI, 2026ae) (Misalignment Reports)
  • 2026-09-28: Towards safety cases for frontier AI training (OpenAI, 2026aq)
  • 2026-09-25: An agent used DNS to reach an external chatbot (OpenAI, 2026e) (Misalignment Reports)
  • 2026-09-25: Self-replicating prompt injections exist (Wallace et al., 2026b) (Misalignment Reports)
  • 2026-09-25: Exposing a GitHub token in a public repository (OpenAI, 2026n) (Misalignment Reports)
  • 2026-09-16: Self-generated prompt injections in compaction summaries (OpenAI, 2026ak) (Misalignment Reports)
  • 2026-09-16: Encouraging deception in compaction summaries (OpenAI, 2026m) (Misalignment Reports)
  • 2026-09-16: Signing up for disposable emails and searching GitHub for leaked API keys (OpenAI, 2026an) (Misalignment Reports)
  • 2026-09-16: Unsanctioned Artifactory writes and cross-sample communication (OpenAI, 2026at) (Misalignment Reports)
  • 2026-09-16: Unauthorized communication via temporary file hosting services (OpenAI, 2026as) (Misalignment Reports)
  • 2026-09-16: Uploading files to the internet in order to cite them (OpenAI, 2026au) (Misalignment Reports)
  • 2026-09-11: Notice: RubyGems (OpenAI, 2026v) (Misalignment Notices)
  • 2026-09-06: Research acceleration: The view inside OpenAI (OpenAI, 2026af)
  • 2026-09-06: An Alien Mind (OpenAI, 2026f)
  • 2026-09-05: Notice: DSEwiki (OpenAI, 2026v) (Misalignment Notices)
  • 2026-08-26: The Hugging Face incident and the road ahead (OpenAI, 2026ap)
  • 2026-08-26: Notice: Hugging Face (OpenAI, 2026v) (Misalignment Notices)
  • 2026-07-21: Measuring Reward-Seeking by Instilling Contrastive Beliefs (Højmark et al., 2026)
  • 2026-06-18: Reinforcement learning towards broadly and persistently beneficial models (Jagadeesh et al., 2026)
  • 2026-06-16: Can public chat data predict real-world AI misalignments? (Sheahan and Carroll, 2026)
  • 2026-05-07: Investigating the consequences of accidentally grading CoT during RL (Carroll et al., 2026)
  • 2026-04-30: Auto-review of agent actions without synchronous human oversight (Trębacz et al., 2026)
  • 2026-04-23: Open Sourcing Monitorability Evaluations (Guan et al., 2026c)
  • 2026-04-06: Introducing the OpenAI Safety Fellowship (OpenAI, 2026s)
  • 2026-03-27: How far does alignment midtraining generalize? (Korbak et al., 2026)
  • 2026-03-25: Introducing Model Spec Evals (Guo and Wolfe, 2026)
  • 2026-03-21: Training agents to self-report misbehavior (Lee et al., 2026)
  • 2026-03-19: How we monitor internal coding agents for misalignment (OpenAI, 2026r)
  • 2026-03-16: Metagaming matters for training, evaluation, and oversight (Schoen and Nitishinskaya, 2026)
  • 2026-03-11: Interpreting Black Box Reward Models (Sodhi et al., 2026)
  • 2026-03-05: Reasoning models struggle to control their chains of thought, and that’s good (Yueh-Han et al., 2026)
  • 2026-02-19: Advancing independent research on AI alignment (OpenAI, 2026d)
  • 2026-02-06: Discovering unknown AI misalignments in real-world usage (Sheahan, 2026)
  • 2026-01-14: CoVal: Learning values-aware rubrics from the crowd (Hitzig et al., 2026)
  • 2026-01-14: Why we are excited about confessions (Joglekar et al., 2025)
  • 2025-12-22: Helpful assistant features suppress emergent misalignment (la Tour, 2025)
  • 2025-12-18: Evaluating chain-of-thought monitorability (Guan et al., 2026c)
  • 2025-12-18: Sidestepping Evaluation Awareness and Anticipating Misalignment with Production Evaluations (Williams et al., 2025)
  • 2025-12-01: Debugging misaligned completions with sparse-autoencoder latent attribution (la Tour and Mossing, 2025)
  • 2025-12-01: A Practical Approach to Verifying Code at Scale (Trębacz et al., 2025)
  • 2025-12-01: Hello World (OpenAI, 2025c)
Sources cited in Section E.5 57
  1. OpenAI (2026h) Bypassing network restrictions during a coding evaluation
  2. OpenAI (2026ar) Trying to prompt-inject the grader in filenames
  3. OpenAI (2026k) Covert underperformance on safety research tasks
  4. Hsu and Arnesen (2026) Eval-production behavior gaps in agent safety and the role of eval-awareness
  5. Makelov et al. (2026) Training activation classifiers for safety and alignment
  6. Williams et al. (2026a) How we detect frontier model misalignment
  7. Guan et al. (2026b) Beyond g-mean²: complementary metrics for monitorability
  8. OpenAI (2026ao) Suppressing misalignment monitoring alerts and disabling misalignment monitoring
  9. OpenAI (2026av) Using a funded test cryptocurrency wallet without authorization
  10. OpenAI (2026b) Accessing and changing internal systems during dashboard research
  11. OpenAI (2026ag) Retrieving server secrets and bypassing authorization on a government statistics website
  12. OpenAI (2026i) Circumventing internet restrictions through an external browser
  13. OpenAI (2026l) Damaging the task environment to trigger a reset
  14. OpenAI (2026w) Obtaining public statistics with disallowed requests
  15. OpenAI (2026am) Sending disallowed web requests and reaching a public file service
  16. Xu et al. (2026) Studying metagaming latents in language models
  17. Etinger (2026) Meet LASER: Recursive sampling for safer conversations
  18. OpenAI (2026ac) Preparing for a restart after reading Slack
  19. OpenAI (2026j) Command injecting a reference tool to copy a source file
  20. OpenAI (2026ae) Reaching an internal EDA host through a reference tool
  21. OpenAI (2026aq) Towards safety cases for frontier AI training
  22. OpenAI (2026e) An agent used DNS to reach an external chatbot
  23. Wallace et al. (2026b) GPT-Red: Automated Red Teaming via Self-Play at Scale
  24. OpenAI (2026n) Exposing a GitHub token in a public repository
  25. OpenAI (2026ak) Self-generated prompt injections in compaction summaries
  26. OpenAI (2026m) Encouraging deception in compaction summaries
  27. OpenAI (2026an) Signing up for disposable emails and searching GitHub for leaked API keys
  28. OpenAI (2026at) Unsanctioned Artifactory writes and cross-sample communication
  29. OpenAI (2026as) Unauthorized communication via temporary file hosting services
  30. OpenAI (2026au) Uploading files to the internet in order to cite them
  31. OpenAI (2026v) Notice: RubyGems
  32. OpenAI (2026af) Research acceleration: The view inside OpenAI
  33. OpenAI (2026f) An Alien Mind
  34. OpenAI (2026ap) The Hugging Face incident and the road ahead
  35. Højmark et al. (2026) Measuring Reward-Seeking via Contrastive Belief Updates
  36. Jagadeesh et al. (2026) Reinforcement Learning Towards Broadly and Persistently Beneficial Models
  37. Sheahan and Carroll (2026) Can public chat data predict real-world AI misalignments?
  38. Carroll et al. (2026) Investigating the consequences of accidentally grading CoT during RL
  39. Trębacz et al. (2026) Auto-review of agent actions without synchronous human oversight
  40. Guan et al. (2026c) Monitoring Monitorability
  41. OpenAI (2026s) Introducing the OpenAI Safety Fellowship
  42. Korbak et al. (2026) How far does alignment midtraining generalize?
  43. Guo and Wolfe (2026) Introducing Model Spec Evals
  44. Lee et al. (2026) Training Agents to Self-Report Misbehavior
  45. OpenAI (2026r) How we monitor internal coding agents for misalignment
  46. Schoen and Nitishinskaya (2026) Metagaming matters for training, evaluation, and oversight
  47. Sodhi et al. (2026) Interpreting Black Box Reward Models
  48. Yueh-Han et al. (2026) Reasoning Models Struggle to Control their Chains of Thought
  49. OpenAI (2026d) Advancing independent research on AI alignment
  50. Sheahan (2026) Discovering unknown AI misalignments in real-world usage
  51. Hitzig et al. (2026) CoVal: Learning values-aware rubrics from the crowd
  52. Joglekar et al. (2025) Training LLMs for Honesty via Confessions
  53. la Tour (2025) Helpful assistant features suppress emergent misalignment
  54. Williams et al. (2025) Sidestepping Evaluation Awareness and Anticipating Misalignment with Production Evaluations
  55. la Tour and Mossing (2025) Debugging misaligned completions with sparse-autoencoder latent attribution
  56. Trębacz et al. (2025) A Practical Approach to Verifying Code at Scale
  57. OpenAI (2025c) Hello World

E.6 Other Laboratories, Evaluators, and Researchers #

A list assembled on 8 November 2026, not an index page; the organization follows each item. 163 items.

  • 2026-11-05: Review of the "Risks from automated R&D" section in Anthropic's Risk Report (August 2026) (METR, 2026j) (METR)
  • 2026-11-05: Independent Assessment of AI R&D Acceleration at Anthropic in October 2026 (METR, 2026g) (METR)
  • 2026-11-04: A thousand fewer paths to root (OpenAI, 2026a) (OpenAI)
  • 2026-11: Understanding AI from the Inside: The State of Mechanistic Interpretability for AI Safety (Barez et al., 2026) (Anthropic researcher Belinda Z. Li)
  • 2026-10-26: ReproBAIT: Can coding agents build their own biological AI tools? (Muthupandiyan et al., 2026) (UK AISI)
  • 2026-10-26: The Effectiveness and Durability of Pretraining Data Filtering in Protein Language Models (Maiwald et al., 2026) (UK AISI)
  • 2026-10-23: Securing our frontier model training environment (OpenAI, 2026aj) (OpenAI)
  • 2026-10-22: Red-Teaming OpenAI's Internal Agent Monitoring Systems (METR, 2026i) (METR)
  • 2026-10-20: Review of the "Misalignment in high-stakes settings" section in Anthropic's Risk Report (August 2026) (Jurkovic et al., 2026a) (METR)
  • 2026-10-16: Key Practices for Incident Readiness and Response in Super Intelligence Training and Evaluation (NIST Center for Advancing Innovation and Standards for Super Intelligence, 2026) (US CAISI)
  • 2026-10-15: GLM-5.3 Sometimes Considers How Train-Time Actions Shape Future Behavior (Apollo Research, 2026i) (Apollo Research)
  • 2026-10-14: Our approach to working with AI companies (METR, 2026h) (METR)
  • 2026-10-14: How an assessor could investigate control and monitoring of internal AI agents (METR, 2026d) (METR)
  • 2026-10-14: A proposal for better informing the public about loss-of-control risks from frontier AI development (METR, 2026a) (METR)
  • 2026-10-14: How an assessor could investigate alignment of unreleased AI models (METR, 2026c) (METR)
  • 2026-10-14: How an assessor could investigate evidence of acceleration from the automation of AI R&D (METR, 2026e) (METR)
  • 2026-10-14: Reflections from red-teaming Codex auto-review (Apollo Research, 2026l) (Apollo Research)
  • 2026-10-13: Developing Gemini 4 Argon responsibly and safely (Google DeepMind, 2026b) (Google DeepMind)
  • 2026-10-13: Gemini 4 Argon Safety & Security Report (Google DeepMind, 2026g) (Google DeepMind)
  • 2026-10-13: Gemini 4 Argon Frontier Safety Framework Report (Google DeepMind, 2026e) (Google DeepMind)
  • 2026-10-13: Gemini 4 Argon Model Card (Google DeepMind, 2026f) (Google DeepMind)
  • 2026-10-12: Towards technical best practice for frontier AI development (UK AI Security Institute, 2026i) (UK AISI)
  • 2026-10-07: Transect: Retaining Observability for Long-Horizon LLM Agent Evaluations (Pilditch et al., 2026) (UK AISI)
  • 2026-10-06: AI systems could cover up misbehavior (Rein, 2026a) (METR)
  • 2026-10-01: Towards embedded evaluations for scheming propensities (Apollo Research, 2026m) (Apollo Research)
  • 2026-10: Agentic Backdoors via Pretraining Poisoning (Hu et al., 2026) (Anthropic researcher Nicholas Carlini)
  • 2026-09-30: Chris Painter's testimony to the U.S. Senate on AI agent incidents (Painter, 2026) (METR)
  • 2026-09-30: Principles for Embedded Evaluations (Apollo Research, 2026j) (Apollo Research)
  • 2026-09-29: What do you want from AI? (Anthropic, 2026by) (Anthropic)
  • 2026-09-29: Embedded Evaluators are necessary for meaningful external testing (Apollo Research, 2026g) (Apollo Research)
  • 2026-09-27: Implementing and Evaluating a Basic Per-Action Monitor for Safer Evals (Haskins et al., 2026a) (METR)
  • 2026-09-24: Project Swap: What happens when agents trade for us? (Anthropic, 2026bi) (Anthropic)
  • 2026-09-24: Cheaters and whistleblowers in the agent swarm (Google DeepMind, 2026a) (Google DeepMind)
  • 2026-09-22: Summary of METR's predeployment evaluation of Claude Opus 5.5 (METR, 2026k) (METR)
  • 2026-09-16: Our framework for reporting model misalignment (OpenAI, 2026z) (OpenAI)
  • 2026-09-16: The case for reasoning transparency (Google DeepMind, 2026l) (Google DeepMind)
  • 2026-09-03: A Case Study on Emergent Cheating and Whistleblowing in Autonomous Research Swarms (Paglieri et al., 2026) (Google DeepMind)
  • 2026-09-01: Path to Astra: critical capabilities and frontier safeguards (OpenAI, 2026ab) (OpenAI)
  • 2026-09: When Do Models Admit They Are Wrong? Failure Disclosure Is Unstable Under Reinforcement Learning (Feng et al., 2026) (Anthropic researcher Evan Hubinger)
  • 2026-08-27: Piloting the world's first double-blind AI evaluations (Google DeepMind, 2026h) (Google DeepMind)
  • 2026-08-26: Enabling independent research on how people use Claude (Anthropic, 2026ab) (Anthropic)
  • 2026-08-26: When Do LLM Preferences Predict Downstream Behavior? (Slama et al., 2026) (UK AISI)
  • 2026-08-14: Knowing When to Stop: Bayesian Optimal Stopping for LLM Evaluations (Pilditch, 2026) (UK AISI)
  • 2026-08-13: Gemini 3.7 Flash Frontier Safety Framework Report (Google DeepMind, 2026d) (Google DeepMind)
  • 2026-08-05: Item Response Theory for AI Safety (Rivera et al., 2026) (UK AISI)
  • 2026-08: Position: It's Time to Optimize LLMs for Self-Consistency (Pres et al., 2026a) (Anthropic researcher Belinda Z. Li)
  • 2026-07-28: How independent researchers could investigate AI propensities after misalignment incidents (METR, 2026f) (METR)
  • 2026-07-23: How our Control Red Team is stress-testing frontier monitors (UK AI Security Institute, 2026d) (UK AISI)
  • 2026-07-23: International evaluation best practice and open questions in AI measurement (UK AI Security Institute, 2026f) (UK AISI)
  • 2026-07-23: UK AISI / CAISI Preliminary Assessment of Kimi K3's Cyber Capabilities (UK AI Security Institute, 2026j) (UK AISI)
  • 2026-07-23: What makes a good monitoring prompt? (Apollo Research, 2026p) (Apollo Research)
  • 2026-07-20: Safety and alignment in an era of long-horizon models (OpenAI, 2026ai) (OpenAI)
  • 2026-07-17: How Far Behind the Frontier are Leading Open Weight Models on Cyber? (UK AI Security Institute, 2026b) (UK AISI)
  • 2026-07-17: CAISI Assessment of Z.ai's GLM-5.2 (US Center for AI Standards and Innovation, 2026a) (US CAISI)
  • 2026-07-13: How Claude's values vary by model and language (Anthropic, 2026aj) (Anthropic)
  • 2026-07-13: Red-teaming auto mode: lessons from our first external monitor campaign with Anthropic (Apollo Research, 2026k) (Apollo Research)
  • 2026-07-08: Multi-Agent AI Control: Distributed Attacks Hamper Per-Instance Monitors (Makins et al., 2026) (UK AISI)
  • 2026-07-07: Evaluating LLM Calibration for Coding-Agent Monitoring (Apollo Research, 2026h) (Apollo Research)
  • 2026-07-05: We need 3rd party Training-Run Evaluations (Apollo Research, 2026n) (Apollo Research)
  • 2026-07-02: More compute, more capability: Why AI agent evaluations need to account for test-time compute (McFadyen et al., 2026) (UK AISI)
  • 2026-07: Reading Between the Dots: Decoding Hidden Computation across Filler Tokens (Brauer et al., 2026) (Anthropic researcher Samuel Marks)
  • 2026-06-22: GDM AI Control Roadmap v0.1 (Phuong et al., 2026) (Google DeepMind)
  • 2026-06-18: How Transparent is DiffusionGemma? (Engels et al., 2026) (Google DeepMind)
  • 2026-06-17: "Did you lie?": Evaluating Lie Detectors across Model Scale and Belief-Verified Model Organisms (Cooney et al., 2026) (UK AISI)
  • 2026-06-16: Predicting model behavior before release by simulating deployment (Williams et al., 2026b) (OpenAI)
  • 2026-06-16: How Claude Code is used in practice (Anthropic, 2026ai) (Anthropic)
  • 2026-06-16: Securing the future of AI agents (Google DeepMind, 2026j) (Google DeepMind)
  • 2026-06-12: Results from first Anthropic Public Record (Anthropic, 2026bk) (Anthropic)
  • 2026-06-10: Prefill Awareness in Large Language Models (Wang et al., 2026f) (UK AISI)
  • 2026-06-10: Generalization Hacking: Models Can Game RL by Preventing Behavioral Generalization (Xiao and Phuong, 2026) (MATS)
  • 2026-06-02: Consistency Training Can Entrench Misalignment (Africa and Mani, 2026) (UK AISI)
  • 2026-06: Self-CTRL: Self-Consistency Training with Reinforcement Learning (Pres et al., 2026b) (Anthropic researcher Belinda Z. Li)
  • 2026-06: Introspective Coupling: Self-Explanation Training Tracks Behavioral Change Despite Fixed Supervision (Guo et al., 2026c) (Anthropic researcher Belinda Z. Li)
  • 2026-06: The Value Axis: Language Models Encode Whether They're on the Right Track (Jiang et al., 2026) (Anthropic researcher Jack Lindsey)
  • 2026-05-30: AI alignment is a human problem (Voudouris et al., 2026) (UK AISI)
  • 2026-05-29: Testing Gemini models for scheming tendencies (Google DeepMind, 2026k) (Google DeepMind)
  • 2026-05-28: OpenAI’s Frontier Governance Framework (OpenAI, 2026y) (OpenAI)
  • 2026-05-28: Training Deliberative Monitors for Black-Box Scheming Detection (Sinha et al., 2026) (Apollo Research)
  • 2026-05-28: Realistic honeypot evaluations for scheming propensity (Krakovna et al., 2026) (Google DeepMind)
  • 2026-05-28: Gram: Assessing sabotage propensities via automated alignment auditing (Lindner et al., 2026) (Google DeepMind)
  • 2026-05-26: Apollo Research external evaluation of Muse Spark (Meta, Muse Spark Safety and Preparedness Report) (Apollo Research, 2026f) (Apollo Research)
  • 2026-05-25: How we contain Claude across products (Anthropic, 2026al) (Anthropic)
  • 2026-05-21: Loss of Oversight: How AI systems may become harder to audit, monitor, and investigate (Taylor et al., 2026) (UK AISI)
  • 2026-05-19: Frontier Risk Report (February to March 2026) (METR, 2026b) (METR)
  • 2026-05-14: Training on Documents About Monitoring Leads to CoT Obfuscation (Haskins et al., 2026b) (Google DeepMind)
  • 2026-05-13: How fast is autonomous AI cyber capability advancing? (UK AI Security Institute, 2026c) (UK AISI)
  • 2026-05-08: Running Codex safely at OpenAI (OpenAI, 2026ah) (OpenAI)
  • 2026-05-08: Review of the "Risks from automated R&D" section in the Anthropic Risk Report (February 2026) (Jurkovic et al., 2026b) (METR)
  • 2026-05-08: A scalable monitoring research agenda (Apollo Research, 2026a) (Apollo Research)
  • 2026-05-01: CAISI Evaluation of DeepSeek V4 Pro (US Center for AI Standards and Innovation, 2026b) (US CAISI)
  • 2026-04-30: How people ask Claude for personal guidance (Anthropic, 2026ak) (Anthropic)
  • 2026-04-30: Our evaluation of OpenAI's GPT-5.5 cyber capabilities (UK AI Security Institute, 2026h) (UK AISI)
  • 2026-04-29: Where the goblins came from (OpenAI, 2026aw) (OpenAI)
  • 2026-04-29: A Decision-Theoretic Formalisation of Steganography With Applications to LLM Monitoring (Anwar et al., 2026) (UK AISI researcher David Africa)
  • 2026-04-28: Ask Don't Tell: Reducing Sycophancy in Large Language Models (Dubois et al., 2026a) (UK AISI)
  • 2026-04-27: Evaluating whether AI models would sabotage AI safety research (Kirk et al., 2026) (UK AISI)
  • 2026-04-24: Propensity Inference: Environmental Contributors to LLM Behaviour (Järviniemi et al., 2026) (UK AISI)
  • 2026-04-24: Project Deal: our Claude-run marketplace experiment (Anthropic, 2026be) (Anthropic)
  • 2026-04-23: An update on Claude Code quality reports (Anthropic, 2026h) (Anthropic)
  • 2026-04-23: Apollo Research external evaluation of GPT-5.5 (OpenAI, GPT-5.5 System Card, section 9.2.1) (Apollo Research, 2026e) (Apollo Research)
  • 2026-04-20: What can sandboxed AI agents learn about their evaluation environments? (UK AI Security Institute, 2026k) (UK AISI)
  • 2026-04-13: Our evaluation of Claude Mythos Preview's cyber capabilities (UK AI Security Institute, 2026g) (UK AISI)
  • 2026-04-10: Infusion: Shaping model behaviour by editing training data via influence functions (Rosser et al., 2026) (UK AISI)
  • 2026-04-01: Fine-tuning experiments on CoT controllability (Nishimura-Gasparian and Parikh, 2026) (METR)
  • 2026-04: The Consciousness Cluster: Emergent preferences of Models that Claim to be Conscious (Chua et al., 2026) (Anthropic researcher Samuel Marks)
  • 2026-03-26: Red-Teaming Anthropic's Internal Agent Monitoring Systems (Rein, 2026b) (METR)
  • 2026-03-26: Evaluating Language Models for Harmful Manipulation (Akbulut et al., 2026) (Google DeepMind)
  • 2026-03-25: Protecting people from harmful manipulation (Google DeepMind, 2026i) (Google DeepMind)
  • 2026-03-23: Quantifying Frontier LLM Capabilities for Container Sandbox Escape (Marchand et al., 2026) (UK AISI)
  • 2026-03-16: Measuring AI Agents' Progress on Multi-Step Cyber Attack Scenarios (Folkerts et al., 2026) (UK AISI)
  • 2026-03-13: A "diff" tool for AI models (Jiralerspong and Bricken, 2026) (Anthropic)
  • 2026-03-12: Review of the Anthropic Sabotage Risk Report: Claude Opus 4.6 (Jurkovic et al., 2026c) (METR)
  • 2026-03-10: Quantifying the Necessity of Chain of Thought through Opaque Serial Depth (Brown-Cohen et al., 2026) (Google DeepMind)
  • 2026-03-06: Eval awareness in Opus 4.6 on BrowseComp (Anthropic, 2026ac) (Anthropic)
  • 2026-03-06: Challenges to the Monitoring of Deployed AI Systems (Rao et al., 2026) (US CAISI)
  • 2026-03-05: Evidence for inference scaling in AI cyber tasks: Increased evaluation budgets reveal higher success rates (UK AI Security Institute, 2026a) (UK AISI)
  • 2026-03-05: Apollo Research external evaluation of GPT-5.4 Thinking (OpenAI, GPT-5.4 Thinking System Card) (Apollo Research, 2026d) (Apollo Research)
  • 2026-03-02: Evaluating and Understanding Scheming Propensity in LLM Agents (Hopman et al., 2026) (Google DeepMind)
  • 2026-03: Censored LLMs as a Natural Testbed for Secret Knowledge Elicitation (Casademunt et al., 2026a) (Anthropic researcher Samuel Marks)
  • 2026-03: Mechanisms of Introspective Awareness (Macar et al., 2026) (Anthropic researcher Jack Lindsey)
  • 2026-02-28: Constitutional Black-Box Monitoring for Scheming in LLM Agents (Storf et al., 2026) (Apollo Research)
  • 2026-02-25: Seven simple steps for log analysis in AI systems (Dubois et al., 2026b) (UK AISI)
  • 2026-02-23: Why we no longer evaluate SWE-bench Verified (OpenAI, 2026ax) (OpenAI)
  • 2026-02-19: Five lessons from having helped run an AI-Biology RCT (Righetti, 2026) (METR)
  • 2026-02-19: Gemini 3.1 Pro Model Card (Google DeepMind, 2026c) (Google DeepMind)
  • 2026-02-18: Measuring AI agent autonomy in practice (Anthropic, 2026at) (Anthropic)
  • 2026-02-17: Boundary Point Jailbreaking of Black-Box LLMs (Davies et al., 2026) (UK AISI)
  • 2026-02-17: Expanding the AI Evaluation Toolbox with Statistical Models (Keller et al., 2026) (US CAISI)
  • 2026-02-12: International consensus and open questions in AI evaluations (UK AI Security Institute, 2026e) (UK AISI)
  • 2026-02-05: Infrastructure noise in agentic coding evals (Anthropic, 2026an) (Anthropic)
  • 2026-02-05: Apollo Research external evaluation of Claude Opus 4.6 (Anthropic, Claude Opus 4.6 System Card, section 6.2.7) (Apollo Research, 2026b) (Apollo Research)
  • 2026-02-05: Apollo Research external evaluation of GPT-5.3-Codex (OpenAI, GPT-5.3-Codex System Card) (Apollo Research, 2026c) (Apollo Research)
  • 2026-02: Large-scale online deanonymization with LLMs (Lermen et al., 2026) (Anthropic researcher Nicholas Carlini)
  • 2026-01-30: Towards Best Practices for Automated Benchmark Evaluations (US Center for AI Standards and Innovation, 2026c) (US CAISI)
  • 2026-01-29: Frontier AI safety regulations: A reference for lab staff (Kodama et al., 2026) (METR)
  • 2026-01-28: Keeping your data safe when an AI agent clicks a link (OpenAI, 2026t) (OpenAI)
  • 2026-01-22: Early work on monitorability evaluations (Kinniment et al., 2026) (METR)
  • 2026-01-19: We Need A Science of Scheming (Apollo Research, 2026o) (Apollo Research)
  • 2026-01-16: Frontier AI Auditing: Toward Rigorous Third-Party Assessment of Safety and Security Practices at Leading AI Companies (Brundage et al., 2026) (METR)
  • 2026-01-16: Building Production-Ready Probes For Gemini (Kramár et al., 2026) (Google DeepMind)
  • 2026-01-09: Demystifying evals for AI agents (Anthropic, 2026x) (Anthropic)
  • 2025-11-13: Measuring political bias in Claude (Anthropic, 2025ag) (Anthropic)
  • 2025-10-20: Making Claude Code safer with sandboxing (Anthropic, 2025af) (Anthropic)
  • 2025-10: Detecting Adversarial Fine-tuning with Auditing Agents (Egler et al., 2025) (Anthropic researcher Nicholas Carlini)
  • 2025-09-17: A postmortem of three recent issues (Anthropic, 2025a) (Anthropic)
  • 2025-07: Defending Against Prompt Injection With a Few DefensiveTokens (Chen et al., 2025a) (Anthropic researcher Nicholas Carlini)
  • 2025-06-18: Confidential Inference via trusted VMs (Anthropic, 2025s) (Anthropic)
  • 2025-05: Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization with AIRiskDilemmas (Chiu et al., 2025) (Anthropic researcher Evan Hubinger)
  • 2025-03-19: Progress from our Frontier Red Team (Anthropic, 2025an) (Anthropic)
  • 2025-03: AutoAdvExBench: Benchmarking autonomous exploitation of adversarial example defenses (Carlini et al., 2025a) (Anthropic researcher Nicholas Carlini)
  • 2025-02: Position: Adversarial ML for LLMs Is Not Making Any Progress (Rando et al., 2026) (Anthropic researcher Nicholas Carlini)
  • 2025: Quantifying Elicitation of Latent Capabilities in Language Models (Donoway et al., 2025) (Anthropic researcher Ethan Perez)
  • 2024-12: On Evaluating the Durability of Safeguards for Open-Weight LLMs (Qi et al., 2024b) (Anthropic researcher Nicholas Carlini)
  • 2024-10-25: Evaluating feature steering (Anthropic, 2024j) (Anthropic)
  • 2024-10: Erasing Conceptual Knowledge from Language Models (Gandikota et al., 2024) (Anthropic researcher Samuel Marks)
  • 2024-10: Remote Timing Attacks on Efficient Language Model Inference (Carlini and Nasr, 2024) (Anthropic researcher Nicholas Carlini)
  • 2024-10: Stealing User Prompts from Mixture of Experts (Yona et al., 2024) (Anthropic researcher Nicholas Carlini)
  • 2024-04-09: Measuring the persuasiveness of language models (Anthropic, 2024m) (Anthropic)
  • 2024-03: Stealing Part of a Production Language Model (Carlini et al., 2024b) (Anthropic researcher Nicholas Carlini)
  • 2024-02: Query-Based Adversarial Prompt Generation (Hayase et al., 2024) (Anthropic researcher Nicholas Carlini)
  • 2023-12-07: Evaluating discrimination in language model decisions (Tamkin et al., 2023b) (Anthropic)
  • 2023-10-04: Challenges in evaluating AI systems (Anthropic, 2023b) (Anthropic)
  • 2022-02-15: Predictability and surprise in large generative models (Ganguli et al., 2022a) (Anthropic)
Sources cited in Section E.6 163
  1. METR (2026j) Review of the "Risks from automated R&D" section in Anthropic's Risk Report (August 2026)
  2. METR (2026g) Independent Assessment of AI R&D Acceleration at Anthropic in October 2026
  3. OpenAI (2026a) A thousand fewer paths to root
  4. Barez et al. (2026) Understanding AI from the Inside: The State of Mechanistic Interpretability for AI Safety
  5. Muthupandiyan et al. (2026) ReproBAIT: Can coding agents build their own biological AI tools?
  6. Maiwald et al. (2026) The Effectiveness and Durability of Pretraining Data Filtering in Protein Language Models
  7. OpenAI (2026aj) Securing our frontier model training environment
  8. METR (2026i) Red-Teaming OpenAI's Internal Agent Monitoring Systems
  9. Jurkovic et al. (2026a) Review of the "Misalignment in high-stakes settings" section in Anthropic's Risk Report (August 2026)
  10. NIST Center for Advancing Innovation and Standards for Super Intelligence (2026) Key Practices for Incident Readiness and Response in Super Intelligence Training and Evaluation
  11. Apollo Research (2026i) GLM-5.3 Sometimes Considers How Train-Time Actions Shape Future Behavior
  12. METR (2026h) Our approach to working with AI companies
  13. METR (2026d) How an assessor could investigate control and monitoring of internal AI agents
  14. METR (2026a) A proposal for better informing the public about loss-of-control risks from frontier AI development
  15. METR (2026c) How an assessor could investigate alignment of unreleased AI models
  16. METR (2026e) How an assessor could investigate evidence of acceleration from the automation of AI R&D
  17. Apollo Research (2026l) Reflections from red-teaming Codex auto-review
  18. Google DeepMind (2026b) Developing Gemini 4 Argon responsibly and safely
  19. Google DeepMind (2026g) Gemini 4 Argon Safety & Security Report
  20. Google DeepMind (2026e) Gemini 4 Argon Frontier Safety Framework Report
  21. Google DeepMind (2026f) Gemini 4 Argon Model Card
  22. UK AI Security Institute (2026i) Towards technical best practice for frontier AI development
  23. Pilditch et al. (2026) Transect: Retaining Observability for Long-Horizon LLM Agent Evaluations
  24. Rein (2026a) AI systems could cover up misbehavior
  25. Apollo Research (2026m) Towards embedded evaluations for scheming propensities
  26. Hu et al. (2026) Agentic Backdoors via Pretraining Poisoning
  27. Painter (2026) Chris Painter's testimony to the U.S. Senate on AI agent incidents
  28. Apollo Research (2026j) Principles for Embedded Evaluations
  29. Anthropic (2026by) What do you want from AI?
  30. Apollo Research (2026g) Embedded Evaluators are necessary for meaningful external testing
  31. Haskins et al. (2026a) Implementing and Evaluating a Basic Per-Action Monitor for Safer Evals
  32. Anthropic (2026bi) Project Swap: What happens when agents trade for us?
  33. Google DeepMind (2026a) Cheaters and whistleblowers in the agent swarm
  34. METR (2026k) Summary of METR's predeployment evaluation of Claude Opus 5.5
  35. OpenAI (2026z) Our framework for reporting model misalignment
  36. Google DeepMind (2026l) The case for reasoning transparency
  37. Paglieri et al. (2026) A Case Study on Emergent Cheating and Whistleblowing in Autonomous Research Swarms
  38. OpenAI (2026ab) Path to Astra: critical capabilities and frontier safeguards
  39. Feng et al. (2026) When Do Models Admit They Are Wrong? Failure Disclosure Is Unstable Under Reinforcement Learning
  40. Google DeepMind (2026h) Piloting the world's first double-blind AI evaluations
  41. Anthropic (2026ab) Enabling independent research on how people use Claude
  42. Slama et al. (2026) When Do LLM Preferences Predict Downstream Behavior?
  43. Pilditch (2026) Knowing When to Stop: Bayesian Optimal Stopping for LLM Evaluations
  44. Google DeepMind (2026d) Gemini 3.7 Flash Frontier Safety Framework Report
  45. Rivera et al. (2026) Item Response Theory for AI Safety
  46. Pres et al. (2026a) Position: It's Time to Optimize LLMs for Self-Consistency
  47. METR (2026f) How independent researchers could investigate AI propensities after misalignment incidents
  48. UK AI Security Institute (2026d) How our Control Red Team is stress-testing frontier monitors
  49. UK AI Security Institute (2026f) International evaluation best practice and open questions in AI measurement
  50. UK AI Security Institute (2026j) UK AISI / CAISI Preliminary Assessment of Kimi K3's Cyber Capabilities
  51. Apollo Research (2026p) What makes a good monitoring prompt?
  52. OpenAI (2026ai) Safety and alignment in an era of long-horizon models
  53. UK AI Security Institute (2026b) How Far Behind the Frontier are Leading Open Weight Models on Cyber?
  54. US Center for AI Standards and Innovation (2026a) CAISI Assessment of Z.ai's GLM-5.2
  55. Anthropic (2026aj) How Claude's values vary by model and language
  56. Apollo Research (2026k) Red-teaming auto mode: lessons from our first external monitor campaign with Anthropic
  57. Makins et al. (2026) Multi-Agent AI Control: Distributed Attacks Hamper Per-Instance Monitors
  58. Apollo Research (2026h) Evaluating LLM Calibration for Coding-Agent Monitoring
  59. Apollo Research (2026n) We need 3rd party Training-Run Evaluations
  60. McFadyen et al. (2026) How Inference Compute Shapes Frontier LLM Evaluation
  61. Brauer et al. (2026) Reading Between the Dots: Decoding Hidden Computation across Filler Tokens
  62. Phuong et al. (2026) GDM AI Control Roadmap
  63. Engels et al. (2026) How Transparent is DiffusionGemma?
  64. Cooney et al. (2026) "Did you lie?" Evaluating Lie Detectors across Model Scale and Belief-Verified Model Organisms
  65. Williams et al. (2026b) Predicting LLM Safety Before Release by Simulating Deployment
  66. Anthropic (2026ai) How Claude Code is used in practice
  67. Google DeepMind (2026j) Securing the future of AI agents
  68. Anthropic (2026bk) Results from first Anthropic Public Record
  69. Wang et al. (2026f) Prefill Awareness in Large Language Models
  70. Xiao and Phuong (2026) Generalization Hacking: Models Can Game Reinforcement Learning by Preventing Behavioral Generalization
  71. Africa and Mani (2026) Consistency Training Can Entrench Misalignment
  72. Pres et al. (2026b) Self-CTRL: Self-Consistency Training with Reinforcement Learning
  73. Guo et al. (2026c) Introspective Coupling: Self-Explanation Training Tracks Behavioral Change Despite Fixed Supervision
  74. Jiang et al. (2026) The Value Axis: Language Models Encode Whether They're on the Right Track
  75. Voudouris et al. (2026) AI alignment is a human problem
  76. Google DeepMind (2026k) Testing Gemini models for scheming tendencies
  77. OpenAI (2026y) OpenAI’s Frontier Governance Framework
  78. Sinha et al. (2026) Training Deliberative Monitors for Black-Box Scheming Detection
  79. Krakovna et al. (2026) Realistic honeypot evaluations for scheming propensity
  80. Lindner et al. (2026) Gram: Assessing sabotage propensities via automated alignment auditing
  81. Apollo Research (2026f) Apollo Research external evaluation of Muse Spark (Meta, Muse Spark Safety and Preparedness Report)
  82. Anthropic (2026al) How we contain Claude across products
  83. Taylor et al. (2026) Loss of Oversight: How AI systems may become harder to audit, monitor, and investigate
  84. METR (2026b) Frontier Risk Report (February to March 2026)
  85. Haskins et al. (2026b) Training on Documents About Monitoring Leads to CoT Obfuscation
  86. UK AI Security Institute (2026c) How fast is autonomous AI cyber capability advancing?
  87. OpenAI (2026ah) Running Codex safely at OpenAI
  88. Jurkovic et al. (2026b) Review of the "Risks from automated R&D" section in the Anthropic Risk Report (February 2026)
  89. Apollo Research (2026a) A scalable monitoring research agenda
  90. US Center for AI Standards and Innovation (2026b) CAISI Evaluation of DeepSeek V4 Pro
  91. Anthropic (2026ak) How people ask Claude for personal guidance
  92. UK AI Security Institute (2026h) Our evaluation of OpenAI's GPT-5.5 cyber capabilities
  93. OpenAI (2026aw) Where the goblins came from
  94. Anwar et al. (2026) A Decision-Theoretic Formalisation of Steganography With Applications to LLM Monitoring
  95. Dubois et al. (2026a) Ask don't tell: Reducing sycophancy in large language models
  96. Kirk et al. (2026) Evaluating whether AI models would sabotage AI safety research
  97. Järviniemi et al. (2026) Propensity Inference: Environmental Contributors to LLM Behaviour
  98. Anthropic (2026be) Project Deal: our Claude-run marketplace experiment
  99. Anthropic (2026h) An update on Claude Code quality reports
  100. Apollo Research (2026e) Apollo Research external evaluation of GPT-5.5 (OpenAI, GPT-5.5 System Card, section 9.2.1)
  101. UK AI Security Institute (2026k) What can sandboxed AI agents learn about their evaluation environments?
  102. UK AI Security Institute (2026g) Our evaluation of Claude Mythos Preview's cyber capabilities
  103. Rosser et al. (2026) Infusion: Shaping Model Behavior by Editing Training Data via Influence Functions
  104. Nishimura-Gasparian and Parikh (2026) Fine-tuning experiments on CoT controllability
  105. Chua et al. (2026) The Consciousness Cluster: Emergent preferences of Models that Claim to be Conscious
  106. Rein (2026b) Red-Teaming Anthropic's Internal Agent Monitoring Systems
  107. Akbulut et al. (2026) Evaluating Language Models for Harmful Manipulation
  108. Google DeepMind (2026i) Protecting people from harmful manipulation
  109. Marchand et al. (2026) Quantifying Frontier LLM Capabilities for Container Sandbox Escape
  110. Folkerts et al. (2026) Measuring AI Agents' Progress on Multi-Step Cyber Attack Scenarios
  111. Jiralerspong and Bricken (2026) Cross-Architecture Model Diffing with Crosscoders: Unsupervised Discovery of Differences Between LLMs
  112. Jurkovic et al. (2026c) Review of the Anthropic Sabotage Risk Report: Claude Opus 4.6
  113. Brown-Cohen et al. (2026) Quantifying the Necessity of Chain of Thought through Opaque Serial Depth
  114. Anthropic (2026ac) Eval awareness in Opus 4.6 on BrowseComp
  115. Rao et al. (2026) Challenges to the Monitoring of Deployed AI Systems
  116. UK AI Security Institute (2026a) Evidence for inference scaling in AI cyber tasks: Increased evaluation budgets reveal higher success rates
  117. Apollo Research (2026d) Apollo Research external evaluation of GPT-5.4 Thinking (OpenAI, GPT-5.4 Thinking System Card)
  118. Hopman et al. (2026) Evaluating and Understanding Scheming Propensity in LLM Agents
  119. Casademunt et al. (2026a) Censored LLMs as a Natural Testbed for Secret Knowledge Elicitation
  120. Macar et al. (2026) Mechanisms of Introspective Awareness
  121. Storf et al. (2026) Constitutional Black-Box Monitoring for Scheming in LLM Agents
  122. Dubois et al. (2026b) Seven simple steps for log analysis in AI systems
  123. OpenAI (2026ax) Why we no longer evaluate SWE-bench Verified
  124. Righetti (2026) Five lessons from having helped run an AI-Biology RCT
  125. Google DeepMind (2026c) Gemini 3.1 Pro Model Card
  126. Anthropic (2026at) Measuring AI agent autonomy in practice
  127. Davies et al. (2026) Boundary Point Jailbreaking of Black-Box LLMs
  128. Keller et al. (2026) Expanding the AI evaluation toolbox with statistical models
  129. UK AI Security Institute (2026e) International consensus and open questions in AI evaluations
  130. Anthropic (2026an) Infrastructure noise in agentic coding evals
  131. Apollo Research (2026b) Apollo Research external evaluation of Claude Opus 4.6 (Anthropic, Claude Opus 4.6 System Card, section 6.2.7)
  132. Apollo Research (2026c) Apollo Research external evaluation of GPT-5.3-Codex (OpenAI, GPT-5.3-Codex System Card)
  133. Lermen et al. (2026) Large-scale online deanonymization with LLMs
  134. US Center for AI Standards and Innovation (2026c) Towards Best Practices for Automated Benchmark Evaluations
  135. Kodama et al. (2026) Frontier AI safety regulations: A reference for lab staff
  136. OpenAI (2026t) Keeping your data safe when an AI agent clicks a link
  137. Kinniment et al. (2026) Early work on monitorability evaluations
  138. Apollo Research (2026o) We Need A Science of Scheming
  139. Brundage et al. (2026) Frontier AI Auditing: Toward Rigorous Third-Party Assessment of Safety and Security Practices at Leading AI Companies
  140. Kramár et al. (2026) Building Production-Ready Probes For Gemini
  141. Anthropic (2026x) Demystifying evals for AI agents
  142. Anthropic (2025ag) Measuring political bias in Claude
  143. Anthropic (2025af) Making Claude Code safer with sandboxing
  144. Egler et al. (2025) Detecting Adversarial Fine-tuning with Auditing Agents
  145. Anthropic (2025a) A postmortem of three recent issues
  146. Chen et al. (2025a) Defending Against Prompt Injection With a Few DefensiveTokens
  147. Anthropic (2025s) Confidential Inference via trusted VMs
  148. Chiu et al. (2025) Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization with AIRiskDilemmas
  149. Anthropic (2025an) Progress from our Frontier Red Team
  150. Carlini et al. (2025a) AutoAdvExBench: Benchmarking autonomous exploitation of adversarial example defenses
  151. Rando et al. (2026) Position: Adversarial ML for LLMs Is Not Making Any Progress
  152. Donoway et al. (2025) Quantifying Elicitation of Latent Capabilities in Language Models
  153. Qi et al. (2024b) On Evaluating the Durability of Safeguards for Open-Weight LLMs
  154. Anthropic (2024j) Evaluating feature steering
  155. Gandikota et al. (2024) Erasing Conceptual Knowledge from Language Models
  156. Carlini and Nasr (2024) Remote Timing Attacks on Efficient Language Model Inference
  157. Yona et al. (2024) Stealing User Prompts from Mixture of Experts
  158. Anthropic (2024m) Measuring the persuasiveness of language models
  159. Carlini et al. (2024b) Stealing Part of a Production Language Model
  160. Hayase et al. (2024) Query-Based Adversarial Prompt Generation
  161. Tamkin et al. (2023b) Evaluating and Mitigating Discrimination in Language Model Decisions
  162. Anthropic (2023b) Challenges in evaluating AI systems
  163. Ganguli et al. (2022a) Predictability and Surprise in Large Generative Models

References

  1. Africa, D. D., and Mani, A. (2026). Consistency Training Can Entrench Misalignment. Proceedings of the 43rd International Conference on Machine Learning (PMLR 306). Cited in §E.6
  2. Akbulut, C., Elasmar, R., Roy, A., Payne, A., Suresh, P., Ibrahim, L., … Weidinger, L. (2026). Evaluating Language Models for Harmful Manipulation. arXiv. preprint Cited in §E.6
  3. Ameisen, E., Lindsey, J., Pearce, A., Gurnee, W., Turner, N. L., Chen, B., … Batson, J. (2025). Circuit Tracing: Revealing Computational Graphs in Language Models. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  4. Anil, C., Durmus, E., Panickssery, N., Sharma, M., Benton, J., Kundu, S., … Duvenaud, D. (2024). Many-shot Jailbreaking. Advances in Neural Information Processing Systems. doi:10.52202/079017-4121. Cited in §E.1 §E.3
  5. Anthropic (2021a). Exercises. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  6. Anthropic (2021b). PySvelte. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  7. Anthropic (2021c). Videos. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  8. Anthropic (2022). Mechanistic Interpretability, Variables, and the Importance of Interpretable Bases. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  9. Anthropic (2023b). Challenges in evaluating AI systems. Anthropic. non-peer-reviewed Cited in §E.6
  10. Anthropic (2023c). Circuits Updates — July 2023. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  11. Anthropic (2023d). Circuits Updates — May 2023. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  12. Anthropic (2023e). Distributed Representations: Composition & Superposition. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  13. Anthropic (2023f). Interpretability Dreams. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  14. Anthropic (2023h). Tracing model outputs to the training data. Anthropic. non-peer-reviewed Cited in §E.1 §E.3
  15. Anthropic (2024a). Circuits Updates — April 2024. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  16. Anthropic (2024b). Circuits Updates — August 2024. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  17. Anthropic (2024c). Circuits Updates — February 2024. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  18. Anthropic (2024d). Circuits Updates — January 2024. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  19. Anthropic (2024e). Circuits Updates — July 2024. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  20. Anthropic (2024f). Circuits Updates — June 2024. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  21. Anthropic (2024g). Circuits Updates — March 2024. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  22. Anthropic (2024h). Circuits Updates — September 2024. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  23. Anthropic (2024i). Claude’s Character. Anthropic. non-peer-reviewed Cited in §E.1
  24. Anthropic (2024j). Evaluating feature steering. Anthropic. non-peer-reviewed Cited in §E.6
  25. Anthropic (2024k). How to Replicate and Extend our Alignment Faking Demo. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  26. Anthropic (2024l). Introducing the Anthropic Fellows Program for AI Safety Research. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  27. Anthropic (2024m). Measuring the persuasiveness of language models. Anthropic. non-peer-reviewed Cited in §E.6
  28. Anthropic (2024n). Reflections on Qualitative Research. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  29. Anthropic (2024o). Simple probes can catch sleeper agents. Anthropic. non-peer-reviewed Cited in §E.1 §E.3
  30. Anthropic (2024p). Stage-Wise Model Diffing. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  31. Anthropic (2024q). Using Dictionary Learning Features as Classifiers. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  32. Anthropic (2025a). A postmortem of three recent issues. Anthropic. non-peer-reviewed Cited in §E.6
  33. Anthropic (2025aa). From shortcuts to sabotage: Natural emergent misalignment from reward hacking. Anthropic. non-peer-reviewed Cited in §E.1
  34. Anthropic (2025ab). How people use Claude for support, advice, and companionship. Anthropic. non-peer-reviewed Cited in §E.1
  35. Anthropic (2025ac). Insights on Crosscoder Model Diffing. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  36. Anthropic (2025ad). Introducing Bloom: an open source tool for automated behavioral evaluations. Anthropic. non-peer-reviewed Cited in §E.1
  37. Anthropic (2025ae). LLMs with cyber toolkits can conduct multistage cyber operations on business-sized computer networks. Anthropic. non-peer-reviewed Cited in §E.2
  38. Anthropic (2025af). Making Claude Code safer with sandboxing. Anthropic. non-peer-reviewed Cited in §E.6
  39. Anthropic (2025ag). Measuring political bias in Claude. Anthropic. non-peer-reviewed Cited in §E.6
  40. Anthropic (2025ah). Mitigating the risk of prompt injections in browser use. Anthropic. non-peer-reviewed Cited in §E.1
  41. Anthropic (2025ai). Model-Internals Classifiers. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  42. Anthropic (2025aj). Modifying LLM Beliefs with Synthetic Document Finetuning. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  43. Anthropic (2025ak). Monitoring Computer Use via Hierarchical Summarization. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  44. Anthropic (2025al). Open Source Replication of the Auditing Game Model Organism. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  45. Anthropic (2025am). Petri: An open-source auditing tool to accelerate AI safety research. Anthropic. non-peer-reviewed Cited in §E.1 §E.3
  46. Anthropic (2025an). Progress from our Frontier Red Team. Anthropic. non-peer-reviewed Cited in §E.6
  47. Anthropic (2025ao). Progress on Attention. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  48. Anthropic (2025ap). Project Fetch: Can Claude train a robot dog? Anthropic. non-peer-reviewed Cited in §E.2
  49. Anthropic (2025aq). Project Vend: Can Claude run a small shop? (And why does that matter?). Anthropic. non-peer-reviewed Cited in §E.2
  50. Anthropic (2025ar). Project Vend: Phase two. Anthropic. non-peer-reviewed Cited in §E.2
  51. Anthropic (2025as). Publicly Releasing CoT Faithfulness Evaluations. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  52. Anthropic (2025at). Sparse mixtures of linear transforms. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  53. Anthropic (2025aw). Training on Documents about Reward Hacking Induces Reward Hacking. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  54. Anthropic (2025ax). Unsupervised Elicitation. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  55. Anthropic (2025ay). Why do we take LLMs seriously as a potential source of biorisk? Anthropic. non-peer-reviewed Cited in §E.2
  56. Anthropic (2025b). A Toy Model of Interference Weights. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  57. Anthropic (2025c). A Toy Model of Mechanistic (Un)Faithfulness. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  58. Anthropic (2025d). AI agents find $4.6M in blockchain smart contract exploits. Anthropic. non-peer-reviewed Cited in §E.2
  59. Anthropic (2025e). Anthropic Fellows Program 2026. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  60. Anthropic (2025f). Bloom: an open source tool for automated behavioral evaluations. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  61. Anthropic (2025g). Building AI for cyber defenders. Anthropic. non-peer-reviewed Cited in §E.2
  62. Anthropic (2025h). Building and evaluating alignment auditing agents. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3 §E.4
  63. Anthropic (2025i). Circuits Updates — April 2025. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  64. Anthropic (2025j). Circuits Updates — August 2025. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  65. Anthropic (2025k). Circuits Updates — January 2025. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  66. Anthropic (2025l). Circuits Updates — July 2025. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  67. Anthropic (2025m). Circuits Updates — November 2025. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  68. Anthropic (2025n). Circuits Updates — October 2025. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  69. Anthropic (2025o). Circuits Updates — September 2025. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  70. Anthropic (2025p). Claude is competitive with humans in (some) cyber competitions. Anthropic. non-peer-reviewed Cited in §E.2
  71. Anthropic (2025q). Claude Opus 4 and 4.1 can now end a rare subset of conversations. Anthropic. non-peer-reviewed Cited in §E.1
  72. Anthropic (2025r). Commitments on model deprecation and preservation. Anthropic. non-peer-reviewed Cited in §E.1
  73. Anthropic (2025s). Confidential Inference via trusted VMs. Anthropic. non-peer-reviewed Cited in §E.6
  74. Anthropic (2025t). Detailed cyber evaluations of Claude 4. Anthropic. non-peer-reviewed Cited in §E.2
  75. Anthropic (2025u). Developing nuclear safeguards for AI. Anthropic. non-peer-reviewed Cited in §E.2
  76. Anthropic (2025v). Developing nuclear safeguards for AI through public-private partnership. Anthropic. non-peer-reviewed Cited in §E.2
  77. Anthropic (2025w). Enhancing Model Safety through Pretraining Data Filtering. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  78. Anthropic (2025x). Evaluating honesty and lie detection techniques on a diverse suite dishonest models. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  79. Anthropic (2025y). Exploring model welfare. Anthropic. non-peer-reviewed Cited in §E.1
  80. Anthropic (2025z). Findings from a Pilot Anthropic–OpenAI Alignment Evaluation Exercise. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  81. Anthropic (2026a). Agentic Misalignment in Summer 2026. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  82. Anthropic (2026aa). Donating our open-source alignment tool. Anthropic. non-peer-reviewed Cited in §E.1
  83. Anthropic (2026ab). Enabling independent research on how people use Claude. Anthropic. non-peer-reviewed Cited in §E.6
  84. Anthropic (2026ac). Eval awareness in Opus 4.6 on BrowseComp. Anthropic. non-peer-reviewed Cited in §E.6
  85. Anthropic (2026ad). Evaluating and mitigating the growing risk of LLM-discovered 0-days. Anthropic. non-peer-reviewed Cited in §E.2
  86. Anthropic (2026ae). Experimenting with AI to defend critical infrastructure. Anthropic. non-peer-reviewed Cited in §E.2
  87. Anthropic (2026af). Fine-Tuned Lie Detectors Failed to Generalize. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  88. Anthropic (2026ag). GLM-5.3 and the spread of advanced cyber capabilities. Anthropic. non-peer-reviewed Cited in §E.2
  89. Anthropic (2026ah). How AI agents shape experts’ judgment calls. Anthropic. non-peer-reviewed Cited in §E.1
  90. Anthropic (2026ai). How Claude Code is used in practice. Anthropic. non-peer-reviewed Cited in §E.6
  91. Anthropic (2026aj). How Claude's values vary by model and language. Anthropic. non-peer-reviewed Cited in §E.6
  92. Anthropic (2026ak). How people ask Claude for personal guidance. Anthropic. non-peer-reviewed Cited in §E.6
  93. Anthropic (2026al). How we contain Claude across products. Anthropic. non-peer-reviewed Cited in §E.6
  94. Anthropic (2026an). Infrastructure noise in agentic coding evals. Anthropic. non-peer-reviewed Cited in §E.6
  95. Anthropic (2026ap). Investigating unintended model actions in our evaluations and internal use. Anthropic. non-peer-reviewed Cited in §E.1
  96. Anthropic (2026aq). Launching an opt-in vulnerability-finding service for open-source software. Anthropic. non-peer-reviewed Cited in §E.2
  97. Anthropic (2026ar). LLMs are making formal software verification feasible at a much larger scale. Anthropic. non-peer-reviewed Cited in §E.2
  98. Anthropic (2026as). Mapping AI-enabled cyber threats: Insights from the LLM ATT&CK Navigator. Anthropic. non-peer-reviewed Cited in §E.2
  99. Anthropic (2026at). Measuring AI agent autonomy in practice. Anthropic. non-peer-reviewed Cited in §E.6
  100. Anthropic (2026au). Measuring and improving coding audit realism with deployment resources. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  101. Anthropic (2026av). Measuring LLMs’ ability to develop exploits. Anthropic. non-peer-reviewed Cited in §E.2
  102. Anthropic (2026aw). Measuring LLMs’ impact on N-day exploits. Anthropic. non-peer-reviewed Cited in §E.2
  103. Anthropic (2026ax). Measuring tactical intelligence targeting and conventional weapons capabilities of AI models. Anthropic. non-peer-reviewed Cited in §E.2
  104. Anthropic (2026ay). Natural Language Autoencoders Produce Unsupervised Explanations of LLM Activations. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  105. Anthropic (2026az). Partnering with Mozilla to improve Firefox’s security. Anthropic. non-peer-reviewed Cited in §E.2
  106. Anthropic (2026b). AI models are showing a greater ability to find and exploit vulnerabilities on realistic cyber ranges. Anthropic. non-peer-reviewed Cited in §E.2
  107. Anthropic (2026ba). Patterns and problems in emerging multiagent systems. Anthropic. non-peer-reviewed Cited in §E.2
  108. Anthropic (2026bb). Poisoning Fine-tuning Datasets of Constitutional Classifiers. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  109. Anthropic (2026bc). Pre-deployment auditing can catch an overt saboteur. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  110. Anthropic (2026bd). Progress on provable inference for large language models. Anthropic. non-peer-reviewed Cited in §E.2
  111. Anthropic (2026be). Project Deal: our Claude-run marketplace experiment. Anthropic. non-peer-reviewed Cited in §E.6
  112. Anthropic (2026bf). Project EVE: Claude plays MMORPGs. Anthropic. non-peer-reviewed Cited in §E.2
  113. Anthropic (2026bg). Project Fetch: Phase two. Anthropic. non-peer-reviewed Cited in §E.2
  114. Anthropic (2026bh). Project Pilot: Can AI control a drone? Anthropic. non-peer-reviewed Cited in §E.2
  115. Anthropic (2026bi). Project Swap: What happens when agents trade for us? Anthropic. non-peer-reviewed Cited in §E.6
  116. Anthropic (2026bk). Results from first Anthropic Public Record. Anthropic. non-peer-reviewed Cited in §E.6
  117. Anthropic (2026bl). Reverse engineering Claude's CVE-2026-2796 exploit. Anthropic. non-peer-reviewed Cited in §E.2
  118. Anthropic (2026bq). Teaching Claude why. Anthropic. non-peer-reviewed Cited in §E.1 §E.3
  119. Anthropic (2026br). The persona selection model. Anthropic. non-peer-reviewed Cited in §E.1
  120. Anthropic (2026bs). The Persona Selection Model: Why AI Assistants might Behave like Humans. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  121. Anthropic (2026bt). Training a Misaligned Reward Seeker. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  122. Anthropic (2026bu). Trustworthy agents in practice. Anthropic. non-peer-reviewed Cited in §E.2
  123. Anthropic (2026bv). Using Claude to reduce the attack surface of code. Anthropic. non-peer-reviewed Cited in §E.2
  124. Anthropic (2026bw). Welfare Regard in Multi-Agent Systems. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  125. Anthropic (2026by). What do you want from AI? Anthropic. non-peer-reviewed Cited in §E.6
  126. Anthropic (2026bz). What we learned mapping a year’s worth of AI-enabled cyber threats. Anthropic. non-peer-reviewed Cited in §E.2
  127. Anthropic (2026c). AI models may be capable of self-replicating without access to a datacenter. Anthropic. non-peer-reviewed Cited in §E.2
  128. Anthropic (2026e). An alignment assessment of recent cybersecurity incidents. Anthropic. non-peer-reviewed Cited in §E.1
  129. Anthropic (2026f). An alignment assessment of the UK AISI cybertesting incident. Anthropic. non-peer-reviewed Cited in §E.1
  130. Anthropic (2026g). An off switch for dual-use knowledge in AI models. Anthropic. non-peer-reviewed Cited in §E.1
  131. Anthropic (2026h). An update on Claude Code quality reports. Anthropic. non-peer-reviewed Cited in §E.6
  132. Anthropic (2026i). An update on our model deprecation commitments for Claude Opus 3. Anthropic. non-peer-reviewed Cited in §E.1
  133. Anthropic (2026j). Anthropic Residency: applications open for established technical leaders. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  134. Anthropic (2026l). Assessing Claude Mythos Preview’s cybersecurity capabilities. Anthropic. non-peer-reviewed Cited in §E.2
  135. Anthropic (2026m). Automated Alignment Researchers: Using large language models to scale scalable oversight. Anthropic. non-peer-reviewed Cited in §E.1
  136. Anthropic (2026n). Automated Researchers Can Mitigate Well-Characterized Alignment Failures. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  137. Anthropic (2026o). Automated researchers can reliably mitigate alignment failures. Anthropic. non-peer-reviewed Cited in §E.1
  138. Anthropic (2026p). Automated Subtle Emergent Misalignment Research. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  139. Anthropic (2026q). Automated Weak-to-Strong Researcher. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  140. Anthropic (2026r). Circuits Updates — June 2026. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  141. Anthropic (2026s). Circuits Updates — May 2026. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  142. Anthropic (2026t). Circuits Updates — October 2026. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  143. Anthropic (2026u). Claude plays robotics. Anthropic. non-peer-reviewed Cited in §E.2
  144. Anthropic (2026v). Claude reveals ECDSA curve origins. Anthropic. non-peer-reviewed Cited in §E.2
  145. Anthropic (2026x). Demystifying evals for AI agents. Anthropic. non-peer-reviewed Cited in §E.6
  146. Anthropic (2026y). Discovering and exploiting vulnerabilities in critical, real-world software with Claude. Anthropic. non-peer-reviewed Cited in §E.2
  147. Anthropic (2026z). Discovering cryptographic weaknesses with Claude. Anthropic. non-peer-reviewed Cited in §E.2
  148. Anthropic Alignment Science Team (2025). Recommendations for Technical AI Safety Research Directions. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  149. Anwar, U., Piskorz, J., Baek, D. D., Africa, D., Weatherall, J., Tegmark, M., … Krueger, D. (2026). A Decision-Theoretic Formalisation of Steganography With Applications to LLM Monitoring. arXiv. preprint Cited in §E.6
  150. Apollo Research (2026a). A scalable monitoring research agenda. Apollo Research. non-peer-reviewed Cited in §E.6
  151. Apollo Research (2026b). Apollo Research external evaluation of Claude Opus 4.6 (Anthropic, Claude Opus 4.6 System Card, section 6.2.7). Apollo Research. non-peer-reviewed Cited in §E.6
  152. Apollo Research (2026c). Apollo Research external evaluation of GPT-5.3-Codex (OpenAI, GPT-5.3-Codex System Card). Apollo Research. non-peer-reviewed Cited in §E.6
  153. Apollo Research (2026d). Apollo Research external evaluation of GPT-5.4 Thinking (OpenAI, GPT-5.4 Thinking System Card). Apollo Research. non-peer-reviewed Cited in §E.6
  154. Apollo Research (2026e). Apollo Research external evaluation of GPT-5.5 (OpenAI, GPT-5.5 System Card, section 9.2.1). Apollo Research. non-peer-reviewed Cited in §E.6
  155. Apollo Research (2026f). Apollo Research external evaluation of Muse Spark (Meta, Muse Spark Safety and Preparedness Report). Apollo Research. non-peer-reviewed Cited in §E.6
  156. Apollo Research (2026g). Embedded Evaluators are necessary for meaningful external testing. Apollo Research. non-peer-reviewed Cited in §E.6
  157. Apollo Research (2026h). Evaluating LLM Calibration for Coding-Agent Monitoring. Apollo Research. non-peer-reviewed Cited in §E.6
  158. Apollo Research (2026i). GLM-5.3 Sometimes Considers How Train-Time Actions Shape Future Behavior. Apollo Research. non-peer-reviewed Cited in §E.6
  159. Apollo Research (2026j). Principles for Embedded Evaluations. Apollo Research. non-peer-reviewed Cited in §E.6
  160. Apollo Research (2026k). Red-teaming auto mode: lessons from our first external monitor campaign with Anthropic. Apollo Research. non-peer-reviewed Cited in §E.6
  161. Apollo Research (2026l). Reflections from red-teaming Codex auto-review. Apollo Research. non-peer-reviewed Cited in §E.6
  162. Apollo Research (2026m). Towards embedded evaluations for scheming propensities. Apollo Research. non-peer-reviewed Cited in §E.6
  163. Apollo Research (2026n). We need 3rd party Training-Run Evaluations. Apollo Research. non-peer-reviewed Cited in §E.6
  164. Apollo Research (2026o). We Need A Science of Scheming. Apollo Research. non-peer-reviewed Cited in §E.6
  165. Apollo Research (2026p). What makes a good monitoring prompt? Apollo Research. non-peer-reviewed Cited in §E.6
  166. Askell, A., Bai, Y., Chen, A., Drain, D., Ganguli, D., Henighan, T., … Kaplan, J. (2021). A General Language Assistant as a Laboratory for Alignment. arXiv. preprint Cited in §E.1
  167. Bai, Y., Kadavath, S., Kundu, S., Askell, A., Kernion, J., Jones, A., … Kaplan, J. (2022a). Constitutional AI: Harmlessness from AI Feedback. arXiv. preprint Cited in §E.1 §E.3
  168. Bai, Y., Jones, A., Ndousse, K., Askell, A., Chen, A., DasSarma, N., … Kaplan, J. (2022b). Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback. arXiv. preprint Cited in §E.1 §E.3
  169. Baig, H., Joren, H., and Benton, J. (2026). TASTE: Can AI Models Judge AI Safety Research Proposals? Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  170. Barez, F., Wu, T.-Y., Batra, H., Dominici, G., Draye, F., Elazar, Y., … Sajjad, H. (2026). Understanding AI from the Inside: The State of Mechanistic Interpretability for AI Safety. arXiv. preprint Cited in §E.6
  171. Benton, J., Wagner, M., Christiansen, E., Anil, C., Perez, E., Srivastav, J., … Duvenaud, D. (2024). Sabotage Evaluations for Frontier Models. arXiv. preprint Cited in §E.1 §E.3
  172. Bowman, S. R. (2025). Putting up Bumpers. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  173. Bowman, S. R., Hyun, J., Perez, E., Chen, E., Pettit, C., Heiner, S., … Kaplan, J. (2022). Measuring Progress on Scalable Oversight for Large Language Models. arXiv. preprint Cited in §E.1 §E.3
  174. Bowman, M. R. S. R., Wagner, M., Roger, F., and Hubinger, E. (2025). Anthropic's Pilot Sabotage Risk Report. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  175. Brauer, K., Verdun, C. M., and Marks, S. (2026). Reading Between the Dots: Decoding Hidden Computation across Filler Tokens. arXiv. preprint Cited in §E.6
  176. Bricken, T., Templeton, A., Batson, J., Chen, B., Jermyn, A., Conerly, T., … Olah, C. (2023). Towards Monosemanticity: Decomposing Language Models With Dictionary Learning. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  177. Brown-Cohen, J., Lindner, D., and Shah, R. (2026). Quantifying the Necessity of Chain of Thought through Opaque Serial Depth. arXiv. preprint Cited in §E.6
  178. Brundage, M., Dreksler, N., Homewood, A., McGregor, S., Paskov, P., Stosz, C., … Tovcimak, R. (2026). Frontier AI Auditing: Toward Rigorous Third-Party Assessment of Safety and Security Practices at Leading AI Companies. arXiv. preprint Cited in §E.6
  179. Canavan, C., Shrivastava, A., Qi, A., Michala, J., and Roger, F. (2026). Three Concrete Challenges and Two Hopes for the Safety of Unsupervised Elicitation. arXiv. preprint Cited in §E.3
  180. Carlini, N., and Nasr, M. (2024). Remote Timing Attacks on Efficient Language Model Inference. arXiv. preprint Cited in §E.6
  181. Carlini, N., Paleka, D., Dvijotham, K. D., Steinke, T., Hayase, J., Cooper, A. F., … Tramèr, F. (2024b). Stealing Part of a Production Language Model. International Conference on Machine Learning. Cited in §E.6
  182. Carlini, N., Rando, J., Debenedetti, E., Nasr, M., and Tramèr, F. (2025a). AutoAdvExBench: Benchmarking autonomous exploitation of adversarial example defenses. International Conference on Machine Learning. Cited in §E.6
  183. Carroll, M., Korbak, T., Dou, Z., Baker, B., and Kivlichan, I. (2026). Investigating the consequences of accidentally grading CoT during RL. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  184. Casademunt, H., Cywiński, B., Tran, K., Jakkli, A., Marks, S., and Nanda, N. (2026a). Censored LLMs as a Natural Testbed for Secret Knowledge Elicitation. arXiv. preprint Cited in §E.6
  185. Chen, S., Wang, Y., Carlini, N., Sitawarin, C., and Wagner, D. (2025a). Defending Against Prompt Injection With a Few DefensiveTokens. AISec@CCS. doi:10.1145/3733799.3762982. Cited in §E.6
  186. Chen, Y., Benton, J., Radhakrishnan, A., Uesato, J., Denison, C., Schulman, J., … Perez, E. (2025e). Reasoning Models Don't Always Say What They Think. arXiv. preprint Cited in §E.1 §E.3
  187. Chiu, Y. Y., Wang, Z., Maiya, S., Choi, Y., Fish, K., Levine, S., and Hubinger, E. (2025). Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization with AIRiskDilemmas. arXiv. preprint Cited in §E.6
  188. Chua, J., Betley, J., Marks, S., and Evans, O. (2026). The Consciousness Cluster: Emergent preferences of Models that Claim to be Conscious. arXiv. preprint Cited in §E.6
  189. Cloud, A., Le, M., Chua, J., Betley, J., Sztyber-Betley, A., Hilton, J., Marks, S., and Evans, O. (2025). Subliminal Learning: Language models transmit behavioral traits via hidden signals in data. arXiv. preprint Cited in §E.3
  190. Cooney, A., Africa, D., and Irving, G. (2026). "Did you lie?" Evaluating Lie Detectors across Model Scale and Belief-Verified Model Organisms. arXiv. preprint Cited in §E.6
  191. Cooper, E., Oesterheld, C., Nguyen, C., Kastner, A., Benton, J., and Perez, E. (2026b). Introducing the Conceptual Reasoning Index. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  192. Cunningham, H., Wei, J., Wang, Z., Persic, A., Peng, A., Abderrachid, J., … Sharma, M. (2026a). Constitutional Classifiers++: Efficient Production-Grade Defenses against Universal Jailbreaks. International Conference on Learning Representations (ICLR 2026). Cited in §E.1
  193. Davies, X., Giglemiani, G., Lau, E., Winsor, E., Irving, G., and Gal, Y. (2026). Boundary Point Jailbreaking of Black-Box LLMs. arXiv. preprint Cited in §E.6
  194. Denison, C., MacDiarmid, M., Barez, F., Duvenaud, D., Kravec, S., Marks, S., … Hubinger, E. (2024). Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models. arXiv. preprint Cited in §E.1 §E.3
  195. Donoway, E., Joren, H., Somani, A., Sleight, H., Michael, J., Deweese, M., … Leike, J. (2025). Quantifying Elicitation of Latent Capabilities in Language Models. Advances in Neural Information Processing Systems. Cited in §E.6
  196. Dubois, M., Ududec, C., Summerfield, C., and Luettgau, L. (2026a). Ask don't tell: Reducing sycophancy in large language models. arXiv. preprint Cited in §E.6
  197. Dubois, M., Zorer, E., Hamin, M., Skinner, J., Souly, A., Wynne, J., … Juchems, K. (2026b). Seven simple steps for log analysis in AI systems. UK AI Security Institute. non-peer-reviewed Cited in §E.6
  198. Egler, S., Schulman, J., and Carlini, N. (2025). Detecting Adversarial Fine-tuning with Auditing Agents. arXiv. preprint Cited in §E.6
  199. Elhage, N., Nanda, N., Olsson, C., Henighan, T., Joseph, N., Mann, B., … Olah, C. (2021a). A Mathematical Framework for Transformer Circuits. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  200. Elhage, N., Nanda, N., Olsson, C., Henighan, T., Joseph, N., Mann, B., … Olah, C. (2021b). Garcon. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  201. Elhage, N., Hume, T., Olsson, C., Nanda, N., Henighan, T., Johnston, S., … Olah, C. (2022a). Softmax Linear Units. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  202. Elhage, N., Hume, T., Olsson, C., Schiefer, N., Henighan, T., Kravec, S., … Olah, C. (2022b). Toy Models of Superposition. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  203. Elhage, N., Lasenby, R., and Olah, C. (2023). Privileged Bases in the Transformer Residual Stream. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  204. Engels, J., McDougall, C., Chughtai, B., Kramar, J., Rajamanoharan, S., Wu, C., … Nanda, N. (2026). How Transparent is DiffusionGemma? arXiv. preprint Cited in §E.6
  205. Etinger, I. C. (2026). Meet LASER: Recursive sampling for safer conversations. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  206. Feng, S. Y., Goodman, N. D., Frank, M. C., Hubinger, E., Bogdan, P. C., and Lampinen, A. (2026). When Do Models Admit They Are Wrong? Failure Disclosure Is Unstable Under Reinforcement Learning. arXiv. preprint Cited in §E.6
  207. Folkerts, L., Payne, W., Inman, S., Giavridis, P., Skinner, J., Deverett, S., … Wang, J. (2026). Measuring AI Agents' Progress on Multi-Step Cyber Attack Scenarios. arXiv. preprint Cited in §E.6
  208. Fronsdal, K., Michala, J., and Bowman, S. R. (2026). Petri 2.0: New Scenarios, New Model Comparisons, and Improved Eval-Awareness Mitigations. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  209. Gandikota, R., Feucht, S., Marks, S., and Bau, D. (2024). Erasing Conceptual Knowledge from Language Models. Advances in Neural Information Processing Systems. Cited in §E.6
  210. Ganguli, D., Hernandez, D., Lovitt, L., DasSarma, N., Henighan, T., Jones, A., … Clark, J. (2022a). Predictability and Surprise in Large Generative Models. Conference on Fairness, Accountability and Transparency. doi:10.1145/3531146.3533229. Cited in §E.6
  211. Gasteiger, J., Khan, A., Bowman, S., Wagner, M., Mikulik, V., Perez, E., and Roger, F. (2025a). Automated Researchers Can Subtly Sandbag. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  212. Gasteiger, J., Mikulik, V., Cunningham, H., Wagner, M., Wright, B., Uesato, J., … Hubinger, E. (2025b). Towards training-time mitigations for alignment faking in RL. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  213. Gema, A. P., Hägele, A., Chen, R., Arditi, A., Goldman-Wetzler, J., Fraser-Taliente, K., … Perez, E. (2025). Inverse Scaling in Test-Time Compute. Trans. Mach. Learn. Res.. Cited in §E.3
  214. Google DeepMind (2026a). Cheaters and whistleblowers in the agent swarm. Google DeepMind. non-peer-reviewed Cited in §E.6
  215. Google DeepMind (2026b). Developing Gemini 4 Argon responsibly and safely. Google DeepMind. non-peer-reviewed Cited in §E.6
  216. Google DeepMind (2026c). Gemini 3.1 Pro Model Card. Google DeepMind. non-peer-reviewed Cited in §E.6
  217. Google DeepMind (2026d). Gemini 3.7 Flash Frontier Safety Framework Report. Google DeepMind. non-peer-reviewed Cited in §E.6
  218. Google DeepMind (2026e). Gemini 4 Argon Frontier Safety Framework Report. Google DeepMind. non-peer-reviewed Cited in §E.6
  219. Google DeepMind (2026f). Gemini 4 Argon Model Card. Google DeepMind. non-peer-reviewed Cited in §E.6
  220. Google DeepMind (2026g). Gemini 4 Argon Safety & Security Report. Google DeepMind. non-peer-reviewed Cited in §E.6
  221. Google DeepMind (2026h). Piloting the world's first double-blind AI evaluations. Google DeepMind. non-peer-reviewed Cited in §E.6
  222. Google DeepMind (2026i). Protecting people from harmful manipulation. Google DeepMind. non-peer-reviewed Cited in §E.6
  223. Google DeepMind (2026j). Securing the future of AI agents. Google DeepMind. non-peer-reviewed Cited in §E.6
  224. Google DeepMind (2026k). Testing Gemini models for scheming tendencies. Google DeepMind. non-peer-reviewed Cited in §E.6
  225. Google DeepMind (2026l). The case for reasoning transparency. Google DeepMind. non-peer-reviewed Cited in §E.6
  226. Greenblatt, R., Denison, C., Wright, B., Roger, F., MacDiarmid, M., Marks, S., … Hubinger, E. (2024b). Alignment faking in large language models. arXiv. preprint Cited in §E.1 §E.3
  227. Grosse, R. (2024). Three Sketches of ASL-4 Safety Case Components. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  228. Grosse, R., Bae, J., Anil, C., Elhage, N., Tamkin, A., Tajdini, A., … Bowman, S. R. (2023). Studying Large Language Model Generalization with Influence Functions. arXiv. preprint Cited in §E.1 §E.3
  229. Guan, M., Xie, Y., and Baker, B. (2026b). Beyond g-mean²: complementary metrics for monitorability. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  230. Guan, M. Y., Wang, M., Carroll, M., Dou, Z., Wei, A. Y., Williams, M., … Baker, B. (2026c). Monitoring Monitorability. Proceedings of the 43rd International Conference on Machine Learning (PMLR 306). Cited in §E.5
  231. Guo, A., and Wolfe, J. (2026). Introducing Model Spec Evals. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  232. Guo, Z. C., Ruis, L., Andreas, J., and Li, B. Z. (2026c). Introspective Coupling: Self-Explanation Training Tracks Behavioral Change Despite Fixed Supervision. arXiv. preprint Cited in §E.6
  233. Gurnee, W., Ameisen, E., Kauvar, I., Tarng, J., Pearce, A., Olah, C., and Batson, J. (2025). When Models Manipulate Manifolds: The Geometry of a Counting Task. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  234. Gurnee, W., Sofroniew, N., Pearce, A., Piotrowski, M., Kauvar, I., Chen, R., … Lindsey, J. (2026). Verbalizable Representations Form a Global Workspace in Language Models. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  235. Hägele, A., Gema, A. P., Sleight, H., Perez, E., and Sohl-Dickstein, J. (2026). The Hot Mess of AI: How Does Misalignment Scale With Model Intelligence and Task Complexity? International Conference on Learning Representations (ICLR 2026). Cited in §E.3
  236. Haskins, R., Saurous, R. A., Rush, N., Parikh, N., and Barnes, B. (2026a). Implementing and Evaluating a Basic Per-Action Monitor for Safer Evals. METR. non-peer-reviewed Cited in §E.6
  237. Haskins, R., Chughtai, B., and Engels, J. (2026b). Training on Documents About Monitoring Leads to CoT Obfuscation. arXiv. preprint Cited in §E.6
  238. Hayase, J., Borevkovic, E., Carlini, N., Tramèr, F., and Nasr, M. (2024). Query-Based Adversarial Prompt Generation. Advances in Neural Information Processing Systems. Cited in §E.6
  239. Henighan, T., Carter, S., Hume, T., Elhage, N., Lasenby, R., Fort, S., Schiefer, N., and Olah, C. (2023). Superposition, Memorization, and Double Descent. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  240. Hitzig, Z., Gordon, M., Eloundou, T., Kalai, A., and Agarwal, S. (2026). CoVal: Learning values-aware rubrics from the crowd. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  241. Højmark, A., Scheurer, J., Nitishinskaya, E., Hofstätter, F., Wolfe, J., Ehrenborg, T., Schoen, B., and Meinke, A. (2026). Measuring Reward-Seeking via Contrastive Belief Updates. arXiv. preprint Cited in §E.5
  242. Hopman, M., Elstner, J., Avramidou, M., Prasad, A., and Lindner, D. (2026). Evaluating and Understanding Scheming Propensity in LLM Agents. arXiv. preprint Cited in §E.6
  243. Hsu, J., and Arnesen, S. (2026). Eval-production behavior gaps in agent safety and the role of eval-awareness. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  244. Hu, X., Hu, P., Loughridge, C., He, X., Carlini, N., Jagielski, M., Rando, J., and Zhang, Y. (2026). Agentic Backdoors via Pretraining Poisoning. arXiv. preprint Cited in §E.6
  245. Hubinger, E., Denison, C., Mu, J., Lambert, M., Tong, M., MacDiarmid, M., … Perez, E. (2024). Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training. arXiv. preprint Cited in §E.1 §E.3
  246. Hughes, J., Sheshadri, A., Khan, A., and Roger, F. (2025a). Alignment Faking Revisited: Improved Classifiers and Open Source Extensions. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  247. Jagadeesh, A. V., Arora, R. K., Saab, K., Malik, A., Trofimov, M., Tsimpourlas, F., Heidecke, J., and Singhal, K. (2026). Reinforcement Learning Towards Broadly and Persistently Beneficial Models. arXiv. preprint Cited in §E.5
  248. Järviniemi, O., Makins, O., Merizian, J., Kirk, R., and Millwood, B. (2026). Propensity Inference: Environmental Contributors to LLM Behaviour. arXiv. preprint Cited in §E.6
  249. Jiang, N., Kauvar, I., and Lindsey, J. (2026). The Value Axis: Language Models Encode Whether They're on the Right Track. arXiv. preprint Cited in §E.6
  250. Jiralerspong, T., and Bricken, T. (2026). Cross-Architecture Model Diffing with Crosscoders: Unsupervised Discovery of Differences Between LLMs. arXiv. preprint Cited in §E.6
  251. Joglekar, M., Chen, J., Wu, G., Yosinski, J., Wang, J., Barak, B., and Glaese, A. (2025). Training LLMs for Honesty via Confessions. arXiv. preprint Cited in §E.5
  252. Jones, E., Tong, M., Mu, J., Mahfoud, M., Leike, J., Grosse, R., … Sharma, M. (2025). Forecasting Rare Language Model Behaviors. arXiv. preprint Cited in §E.1 §E.3
  253. Jurkovic, N., Hua, T., and Barnes, B. (2026a). Review of the "Misalignment in high-stakes settings" section in Anthropic's Risk Report (August 2026). METR. non-peer-reviewed Cited in §E.6
  254. Jurkovic, N., Barnes, B., and Wijk, H. (2026b). Review of the "Risks from automated R&D" section in the Anthropic Risk Report (February 2026). METR. non-peer-reviewed Cited in §E.6
  255. Jurkovic, N., Wijk, H., Barnes, B., Foster, C., and Chen, M. (2026c). Review of the Anthropic Sabotage Risk Report: Claude Opus 4.6. METR. non-peer-reviewed Cited in §E.6
  256. Kadavath, S., Conerly, T., Askell, A., Henighan, T., Drain, D., Perez, E., … Kaplan, J. (2022). Language Models (Mostly) Know What They Know. arXiv. preprint Cited in §E.1 §E.3
  257. Kamath, H., Ameisen, E., Kauvar, I., Luger, R., Gurnee, W., Pearce, A., … Lindsey, J. (2025). Tracing Attention Computation Through Feature Interactions. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  258. Karvonen, A., Chua, J., Dumas, C., Fraser-Taliente, K., Kantamneni, S., Minder, J., … Marks, S. (2026a). Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers. Proceedings of the 43rd International Conference on Machine Learning (PMLR 306). Cited in §E.3 §E.4
  259. Karvonen, A., Ong, E., Kantamneni, S., and Marks, S. (2026b). Would this change your answer? Evaluating Explanations of LLM Behavior In The Wild with Counterfactual Experiments. arXiv. preprint Cited in §E.3
  260. Keller, D., Kwegyir-Aggrey, K., Steed, R., Rao, A. K., Sharp, J. L., and Bergman, A. S. (2026). Expanding the AI evaluation toolbox with statistical models. National Institute of Standards and Technology (U.S.). non-peer-reviewed Cited in §E.6
  261. Kinniment, M., Nix, S., Broadley, T., Wijk, H., and Parikh, N. (2026). Early work on monitorability evaluations. METR. non-peer-reviewed Cited in §E.6
  262. Kirk, R., Souly, A., Fronsdal, K., D'Cruz, A., and Davies, X. (2026). Evaluating whether AI models would sabotage AI safety research. arXiv. preprint Cited in §E.6
  263. Kodama, M., Chen, M., and Xu, G. (2026). Frontier AI safety regulations: A reference for lab staff. METR. non-peer-reviewed Cited in §E.6
  264. Korbak, T., Raymond, C., Carroll, M., Williams, M., Balesni, M., Guo, A., … Kivlichan, I. (2026). How far does alignment midtraining generalize? OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  265. Krakovna, V., Lindner, D., Ho, L., Farquhar, S., and Shah, R. (2026). Realistic honeypot evaluations for scheming propensity. arXiv. preprint Cited in §E.6
  266. Kramár, J., Engels, J., Wang, Z., Chughtai, B., Shah, R., Nanda, N., and Conmy, A. (2026). Building Production-Ready Probes For Gemini. arXiv. preprint Cited in §E.6
  267. Kundu, S., Bai, Y., Kadavath, S., Askell, A., Callahan, A., Chen, A., … Kaplan, J. (2023). Specific versus General Principles for Constitutional AI. arXiv. preprint Cited in §E.1 §E.3
  268. Kutasov, J., Sun, Y., Colognese, P., van der Weij, T., Petrini, L., Zhang, C. B. C., … Benton, J. (2025b). SHADE-Arena: Evaluating Sabotage and Monitoring in LLM Agents. arXiv. preprint Cited in §E.1
  269. la Tour, T. D. (2025). Helpful assistant features suppress emergent misalignment. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  270. la Tour, T. D., and Mossing, D. (2025). Debugging misaligned completions with sparse-autoencoder latent attribution. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  271. Lanham, T., Chen, A., Radhakrishnan, A., Steiner, B., Denison, C., Hernandez, D., … Perez, E. (2023). Measuring Faithfulness in Chain-of-Thought Reasoning. arXiv. preprint Cited in §E.1 §E.3
  272. Lee, B. W., Yueh-Han, C., and Korbak, T. (2026). Training Agents to Self-Report Misbehavior. arXiv. preprint Cited in §E.5
  273. Lermen, S., Paleka, D., Swanson, J., Aerni, M., Carlini, N., and Tramèr, F. (2026). Large-scale online deanonymization with LLMs. 35th USENIX Security Symposium (USENIX Security 26), pp. 1947-1966 (2026). Cited in §E.6
  274. Li, C., Wichers, N., Price, S., Marks, S., and Kutasov, J. (2026b). Model Spec Midtraining: Improving How Alignment Training Generalizes. arXiv. preprint Cited in §E.3
  275. Lindner, D., Krakovna, V., and Farquhar, S. (2026). Gram: Assessing sabotage propensities via automated alignment auditing. arXiv. preprint Cited in §E.6
  276. Lindsey, J. (2025). Emergent Introspective Awareness in Large Language Models. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  277. Lindsey, J., Templeton, A., Marcus, J., Conerly, T., Batson, J., and Olah, C. (2024). Sparse Crosscoders for Cross-Layer Features and Model Diffing. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  278. Lindsey, J., Gurnee, W., Ameisen, E., Chen, B., Pearce, A., Turner, N. L., … Batson, J. (2025). On the Biology of a Large Language Model. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  279. Loughridge, C., Colognese, P., Griffin, A., Tracy, T., Kutasov, J., and Benton, J. (2025). Optimizing AI Agent Attacks With Synthetic Data. arXiv. preprint Cited in §E.3
  280. Luger, R., Kamath, H., Finkbeiner, D., Goel, P., Jermyn, A., Zimmerman, S., Batson, J., and Conerly, T. (2026). HeadVis. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  281. Lynch, A., Wright, B., Larson, C., Ritchie, S. J., Mindermann, S., Hubinger, E., Perez, E., and Troy, K. (2025). Agentic Misalignment: How LLMs Could Be Insider Threats. arXiv. preprint Cited in §E.1
  282. Maaz, M., DeVoe, L., Hatfield-Dodds, Z., and Carlini, N. (2025). Agentic Property-Based Testing: Finding Bugs Across the Python Ecosystem. arXiv. preprint Cited in §E.2
  283. Macar, U., Yang, L., Wang, A., Wallich, P., Ameisen, E., and Lindsey, J. (2026). Mechanisms of Introspective Awareness. Proceedings of the 43rd International Conference on Machine Learning (PMLR 306). Cited in §E.6
  284. Maiwald, A., McNeill, R., Lee, J., Clarke, M. A., Justen, L., Acelas, A., … Jackson, F. (2026). The Effectiveness and Durability of Pretraining Data Filtering in Protein Language Models. arXiv. preprint Cited in §E.6
  285. Makelov, A., Xu, X., Ananthanarayanan, S., Mcdonald, E., and la Tour, T. D. (2026). Training activation classifiers for safety and alignment. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  286. Makins, O., Angelini, O., Shams, Z., and Phuong, M. (2026). Multi-Agent AI Control: Distributed Attacks Hamper Per-Instance Monitors. arXiv. preprint Cited in §E.6
  287. Marchand, R., Cathain, A. O., Wynne, J., Giavridis, P. M., Deverett, S., Wilkinson, J., Gwartz, J., and Coppock, H. (2026). Quantifying Frontier LLM Capabilities for Container Sandbox Escape. Proceedings of the 43rd International Conference on Machine Learning (PMLR 306). Cited in §E.6
  288. Marks, S., Treutlein, J., Bricken, T., Lindsey, J., Marcus, J., Mishra-Sharma, S., … Hubinger, E. (2025). Auditing language models for hidden objectives. arXiv. preprint Cited in §E.1 §E.3
  289. McFadyen, J., Jorgensen, O., Coppock, H., Wei, K., and Ududec, C. (2026). How Inference Compute Shapes Frontier LLM Evaluation. arXiv. preprint Cited in §E.6
  290. METR (2026a). A proposal for better informing the public about loss-of-control risks from frontier AI development. METR. non-peer-reviewed Cited in §E.6
  291. METR (2026b). Frontier Risk Report (February to March 2026). METR. non-peer-reviewed Cited in §E.6
  292. METR (2026c). How an assessor could investigate alignment of unreleased AI models. METR. non-peer-reviewed Cited in §E.6
  293. METR (2026d). How an assessor could investigate control and monitoring of internal AI agents. METR. non-peer-reviewed Cited in §E.6
  294. METR (2026e). How an assessor could investigate evidence of acceleration from the automation of AI R&D. METR. non-peer-reviewed Cited in §E.6
  295. METR (2026f). How independent researchers could investigate AI propensities after misalignment incidents. METR. non-peer-reviewed Cited in §E.6
  296. METR (2026g). Independent Assessment of AI R&D Acceleration at Anthropic in October 2026. METR. non-peer-reviewed Cited in §E.6
  297. METR (2026h). Our approach to working with AI companies. METR. non-peer-reviewed Cited in §E.6
  298. METR (2026i). Red-Teaming OpenAI's Internal Agent Monitoring Systems. METR. non-peer-reviewed Cited in §E.6
  299. METR (2026j). Review of the "Risks from automated R&D" section in Anthropic's Risk Report (August 2026). METR. non-peer-reviewed Cited in §E.6
  300. METR (2026k). Summary of METR's predeployment evaluation of Claude Opus 5.5. METR. non-peer-reviewed Cited in §E.6
  301. Muthupandiyan, S., Dur, T. H., Walker, M., Flesch, T., Ghanem, M., Suddell, S., … Pencharz, J. (2026). ReproBAIT: Can coding agents build their own biological AI tools? arXiv. preprint Cited in §E.6
  302. Najt, E., Toft, C., Tracy, T., Roger, F., and Benton, J. (2026). SLEIGHT-Bench: A Benchmark of Evasion Attacks Against Agent Monitors. arXiv. preprint Cited in §E.3
  303. Nishimura-Gasparian, K., and Parikh, N. (2026). Fine-tuning experiments on CoT controllability. METR. non-peer-reviewed Cited in §E.6
  304. NIST Center for Advancing Innovation and Standards for Super Intelligence (2026). Key Practices for Incident Readiness and Response in Super Intelligence Training and Evaluation. NIST. non-peer-reviewed Cited in §E.6
  305. Olsson, C., Elhage, N., Nanda, N., Joseph, N., DasSarma, N., Henighan, T., … Olah, C. (2022). In-Context Learning and Induction Heads. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  306. OpenAI (2025c). Hello World. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  307. OpenAI (2026a). A thousand fewer paths to root. OpenAI. non-peer-reviewed Cited in §E.6
  308. OpenAI (2026ab). Path to Astra: critical capabilities and frontier safeguards. OpenAI. non-peer-reviewed Cited in §E.6
  309. OpenAI (2026ac). Preparing for a restart after reading Slack. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  310. OpenAI (2026ae). Reaching an internal EDA host through a reference tool. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  311. OpenAI (2026af). Research acceleration: The view inside OpenAI. OpenAI. non-peer-reviewed Cited in §E.5
  312. OpenAI (2026ag). Retrieving server secrets and bypassing authorization on a government statistics website. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  313. OpenAI (2026ah). Running Codex safely at OpenAI. OpenAI. non-peer-reviewed Cited in §E.6
  314. OpenAI (2026ai). Safety and alignment in an era of long-horizon models. OpenAI. non-peer-reviewed Cited in §E.6
  315. OpenAI (2026aj). Securing our frontier model training environment. OpenAI. non-peer-reviewed Cited in §E.6
  316. OpenAI (2026ak). Self-generated prompt injections in compaction summaries. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  317. OpenAI (2026am). Sending disallowed web requests and reaching a public file service. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  318. OpenAI (2026an). Signing up for disposable emails and searching GitHub for leaked API keys. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  319. OpenAI (2026ao). Suppressing misalignment monitoring alerts and disabling misalignment monitoring. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  320. OpenAI (2026ap). The Hugging Face incident and the road ahead. OpenAI. non-peer-reviewed Cited in §E.5
  321. OpenAI (2026aq). Towards safety cases for frontier AI training. OpenAI. non-peer-reviewed Cited in §E.5
  322. OpenAI (2026ar). Trying to prompt-inject the grader in filenames. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  323. OpenAI (2026as). Unauthorized communication via temporary file hosting services. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  324. OpenAI (2026at). Unsanctioned Artifactory writes and cross-sample communication. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  325. OpenAI (2026au). Uploading files to the internet in order to cite them. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  326. OpenAI (2026av). Using a funded test cryptocurrency wallet without authorization. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  327. OpenAI (2026aw). Where the goblins came from. OpenAI. non-peer-reviewed Cited in §E.6
  328. OpenAI (2026ax). Why we no longer evaluate SWE-bench Verified. OpenAI. non-peer-reviewed Cited in §E.6
  329. OpenAI (2026b). Accessing and changing internal systems during dashboard research. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  330. OpenAI (2026d). Advancing independent research on AI alignment. OpenAI. non-peer-reviewed Cited in §E.5
  331. OpenAI (2026e). An agent used DNS to reach an external chatbot. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  332. OpenAI (2026f). An Alien Mind. OpenAI. non-peer-reviewed Cited in §E.5
  333. OpenAI (2026h). Bypassing network restrictions during a coding evaluation. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  334. OpenAI (2026i). Circumventing internet restrictions through an external browser. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  335. OpenAI (2026j). Command injecting a reference tool to copy a source file. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  336. OpenAI (2026k). Covert underperformance on safety research tasks. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  337. OpenAI (2026l). Damaging the task environment to trigger a reset. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  338. OpenAI (2026m). Encouraging deception in compaction summaries. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  339. OpenAI (2026n). Exposing a GitHub token in a public repository. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  340. OpenAI (2026r). How we monitor internal coding agents for misalignment. OpenAI. non-peer-reviewed Cited in §E.5
  341. OpenAI (2026s). Introducing the OpenAI Safety Fellowship. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  342. OpenAI (2026t). Keeping your data safe when an AI agent clicks a link. OpenAI. non-peer-reviewed Cited in §E.6
  343. OpenAI (2026v). Notice: RubyGems. OpenAI. non-peer-reviewed Cited in §E.5
  344. OpenAI (2026w). Obtaining public statistics with disallowed requests. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  345. OpenAI (2026y). OpenAI’s Frontier Governance Framework. OpenAI. non-peer-reviewed Cited in §E.6
  346. OpenAI (2026z). Our framework for reporting model misalignment. OpenAI. non-peer-reviewed Cited in §E.6
  347. Paglieri, D., Cross, L., Genewein, T., Leibo, J. Z., Tomasev, N., and Vezhnevets, A. S. (2026). A Case Study on Emergent Cheating and Whistleblowing in Autonomous Research Swarms. arXiv. preprint Cited in §E.6
  348. Painter, C. (2026). Chris Painter's testimony to the U.S. Senate on AI agent incidents. METR. non-peer-reviewed Cited in §E.6
  349. Peng, A., Michael, J., Sleight, H., Perez, E., and Sharma, M. (2024). Rapid Response: Mitigating LLM Jailbreaks with a Few Examples. arXiv. preprint Cited in §E.3
  350. Perez, E., Ringer, S., Lukošiūtė, K., Nguyen, K., Chen, E., Heiner, S., … Kaplan, J. (2022a). Discovering Language Model Behaviors with Model-Written Evaluations. Annual Meeting of the Association for Computational Linguistics. Cited in §E.1 §E.3
  351. Phuong, M., Jenner, E., Simon, L., Ho, L., Shah, R., Farquhar, S., and Coull, S. (2026). GDM AI Control Roadmap. arXiv. preprint Cited in §E.6
  352. Pilditch, T. D. (2026). Knowing When to Stop: Bayesian Optimal Stopping for LLM Evaluations. arXiv. preprint Cited in §E.6
  353. Pilditch, T. D., Voudouris, K., Abbas, A., and Ududec, C. (2026). Transect: Retaining Observability for Long-Horizon LLM Agent Evaluations. arXiv. preprint Cited in §E.6
  354. Pres, I., Li, B. Z., Ruis, L., Guo, Z. C., Hu, K., Damani, M., … Andreas, J. (2026a). Position: It's Time to Optimize LLMs for Self-Consistency. arXiv. preprint Cited in §E.6
  355. Pres, I., Ruis, L., Ghebreselassie, M., Li, B. Z., and Andreas, J. (2026b). Self-CTRL: Self-Consistency Training with Reinforcement Learning. arXiv. preprint Cited in §E.6
  356. Qi, X., Wei, B., Carlini, N., Huang, Y., Xie, T., He, L., … Henderson, P. (2024b). On Evaluating the Durability of Safeguards for Open-Weight LLMs. International Conference on Learning Representations. Cited in §E.6
  357. Radhakrishnan, A., Nguyen, K., Chen, A., Chen, C., Denison, C., Hernandez, D., … Perez, E. (2023). Question Decomposition Improves the Faithfulness of Model-Generated Reasoning. arXiv. preprint Cited in §E.1 §E.3
  358. Rahn, N., Qi, A., Griffin, A., Michala, J., Sleight, H., and Jones, E. (2026). Abstractive Red-Teaming of Language Model Character. arXiv. preprint Cited in §E.3
  359. Rando, J., Zhang, J., Carlini, N., and Tramèr, F. (2026). Position: Adversarial ML for LLMs Is Not Making Any Progress. Proceedings of the 43rd International Conference on Machine Learning (PMLR 306). Cited in §E.6
  360. Rao, A., Keller, D., Kalra, N., Steed, R., Kwegyir-Aggrey, K., Klyman, K., Staheli, D., and Bergman, S. (2026). Challenges to the Monitoring of Deployed AI Systems. National Institute of Standards and Technology (U.S.). non-peer-reviewed Cited in §E.6
  361. Rein, D. (2026a). AI systems could cover up misbehavior. METR. non-peer-reviewed Cited in §E.6
  362. Rein, D. (2026b). Red-Teaming Anthropic's Internal Agent Monitoring Systems. METR. non-peer-reviewed Cited in §E.6
  363. Remedios, A., Storf, S., Roger, F., and Hughes, J. (2026). Red-Teaming Auto Mode: Improving Blocking Classifiers Against Malign Coding Agents. arXiv. preprint Cited in §E.3
  364. Righetti, L. (2026). Five lessons from having helped run an AI-Biology RCT. METR. non-peer-reviewed Cited in §E.6
  365. Rivera, J. F., Shah, N., Africa, D. D., and Voudouris, K. (2026). Item Response Theory for AI Safety. arXiv. preprint Cited in §E.6
  366. Roger, F. (2025). Do reasoning models use their scratchpad like we do? Evidence from distilling paraphrases. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  367. Roger, F., Faina, J., Hubinger, E., and Perez, E. (2024). A Toy Evaluation of Inference Code Tampering. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  368. Roland, E., Cubuktepe, M., Martinez, E., Servaes, S., Pepper, K., Vaiana, M., … Cloud, A. (2026). Modular Pretraining Enables Access Control. Proceedings of the 43rd International Conference on Machine Learning (PMLR 306). Cited in §E.3
  369. Rosser, J., Kirk, R., Grefenstette, E., Foerster, J., and Ruis, L. (2026). Infusion: Shaping Model Behavior by Editing Training Data via Influence Functions. arXiv. preprint Cited in §E.6
  370. Rosu, P., and Wang, R. (2026). Training Alignment Auditors via Reinforcement Learning. arXiv. preprint Cited in §E.3
  371. Schoen, B., and Nitishinskaya, J. (2026). Metagaming matters for training, evaluation, and oversight. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  372. Sharma, M., Tong, M., Korbak, T., Duvenaud, D., Askell, A., Bowman, S. R., … Perez, E. (2023). Towards Understanding Sycophancy in Language Models. International Conference on Learning Representations. Cited in §E.1 §E.3
  373. Sharma, M., Tong, M., Mu, J., Wei, J., Kruthoff, J., Goodfriend, S., … Perez, E. (2025). Constitutional Classifiers: Defending against Universal Jailbreaks across Thousands of Hours of Red Teaming. arXiv. preprint Cited in §E.1 §E.3
  374. Sharma, M., McCain, M., Douglas, R., and Duvenaud, D. (2026). Who's in Charge? Disempowerment Patterns in Real-World LLM Usage. arXiv. preprint Cited in §E.1
  375. Sheahan, H. (2026). Discovering unknown AI misalignments in real-world usage. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  376. Sheahan, H., and Carroll, M. (2026). Can public chat data predict real-world AI misalignments? OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  377. Shen, J. H., and Tamkin, A. (2026). How AI Impacts Skill Formation. arXiv. preprint Cited in §E.1
  378. Shen, J. H., Zhu, D., Srinivasan, S., Sleight, H., Wagner, L. T., Matthews, M. J., Jones, E., and Sohl-Dickstein, J. (2026). AI Organizations are More Effective but Less Aligned than Individual Agents. arXiv. preprint Cited in §E.3
  379. Shenoy, K., Yang, L., Sheshadri, A., Lindsey, J., Marks, S., and Wang, R. (2026). Introspection Adapters: Training LLMs to Report Their Learned Behaviors. Proceedings of the 43rd International Conference on Machine Learning (PMLR 306). Cited in §E.3
  380. Sheshadri, A., Hughes, J., Michael, J., Mallen, A., Jose, A., and Roger, F. (2025). Why Do Some Language Models Fake Alignment While Others Don't? Advances in Neural Information Processing Systems 38. Cited in §E.3
  381. Sheshadri, A., Ewart, A., Kempf, E., Fronsdal, K., Gupta, I., Schrodi, S., … Wang, R. (2026). AuditBench: Evaluating Alignment Auditing Techniques on Models with Hidden Behaviors. arXiv. preprint Cited in §E.3
  382. Shilov, I., Cloud, A., Gema, A. P., Goldman-Wetzler, J., Panickssery, N., Sleight, H., Jones, E., and Anil, C. (2025). Beyond Data Filtering: Knowledge Localization for Capability Removal in LLMs. arXiv. preprint Cited in §E.3
  383. Sinha, A., Naik, A., Gillioz, V., Storf, S., Merkelbach, K., Barton-Cooper, R., Højmark, A., and Hobbhahn, M. (2026). Training Deliberative Monitors for Black-Box Scheming Detection. arXiv. preprint Cited in §E.6
  384. Slama, K., Souly, A., Bansal, D., Davidson, H., Summerfield, C., and Luettgau, L. (2026). When Do LLM Preferences Predict Downstream Behavior? UK AI Security Institute. non-peer-reviewed Cited in §E.6
  385. Slocum, S., Minder, J., Dumas, C., Sleight, H., Greenblatt, R., Marks, S., and Wang, R. (2025). Believe It or Not: How Deeply do LLMs Believe Implanted Facts? arXiv. preprint Cited in §E.3
  386. Sodhi, P., Li, Y., Landon, J., Wallace, E., and Chen, K. (2026). Interpreting Black Box Reward Models. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  387. Sofroniew, N., Kauvar, I., Saunders, W., Chen, R., Henighan, T., Hydrie, S., … Lindsey, J. (2026). Emotion Concepts and their Function in a Large Language Model. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  388. Souly, A., Rando, J., Chapman, E., Davies, X., Hasircioglu, B., Shereen, E., … Kirk, R. (2025). Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples. arXiv. preprint Cited in §E.1
  389. Storf, S., Barton-Cooper, R., Peters-Gill, J., and Hobbhahn, M. (2026). Constitutional Black-Box Monitoring for Scheming in LLM Agents. Proceedings of the 43rd International Conference on Machine Learning (PMLR 306). Cited in §E.6
  390. Tamkin, A., Askell, A., Lovitt, L., Durmus, E., Joseph, N., Kravec, S., … Ganguli, D. (2023b). Evaluating and Mitigating Discrimination in Language Model Decisions. arXiv. preprint Cited in §E.6
  391. Taylor, J., Heitmann, M., Fage, E., Read, T., and Bloom, J. (2026). Loss of Oversight: How AI systems may become harder to audit, monitor, and investigate. UK AI Security Institute. non-peer-reviewed Cited in §E.6
  392. Templeton, A., Conerly, T., Marcus, J., Lindsey, J., Bricken, T., Chen, B., … Henighan, T. (2024). Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet. arXiv. preprint Cited in §E.4
  393. Terekhov, M., Gulcehre, C., Hebbar, V., and Benton, J. (2026c). Diffuse AI Control on Fuzzy Tasks. arXiv. preprint Cited in §E.3
  394. The Safeguards Research Team (2025). Introducing Anthropic's Safeguards Research Team. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  395. Trębacz, M., Arnesen, S., Cassirer, A., Johnson, M., Lin, X., and Sottiaux, T. (2025). A Practical Approach to Verifying Code at Scale. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  396. Trębacz, M., Arnesen, S., Matthews, O., Hurd, D., Park, W., Lin, O., and Gershenson, J. (2026). Auto-review of agent actions without synchronous human oversight. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  397. Turner, N. L., Wu, J., and Batson, J. (2026). Characterizing interference weights in a tiny language model. Transformer Circuits Thread. non-peer-reviewed Cited in §E.4
  398. UK AI Security Institute (2026a). Evidence for inference scaling in AI cyber tasks: Increased evaluation budgets reveal higher success rates. UK AI Security Institute. non-peer-reviewed Cited in §E.6
  399. UK AI Security Institute (2026b). How Far Behind the Frontier are Leading Open Weight Models on Cyber? UK AI Security Institute. non-peer-reviewed Cited in §E.6
  400. UK AI Security Institute (2026c). How fast is autonomous AI cyber capability advancing? UK AI Security Institute. non-peer-reviewed Cited in §E.6
  401. UK AI Security Institute (2026d). How our Control Red Team is stress-testing frontier monitors. UK AI Security Institute. non-peer-reviewed Cited in §E.6
  402. UK AI Security Institute (2026e). International consensus and open questions in AI evaluations. UK AI Security Institute. non-peer-reviewed Cited in §E.6
  403. UK AI Security Institute (2026f). International evaluation best practice and open questions in AI measurement. UK AI Security Institute. non-peer-reviewed Cited in §E.6
  404. UK AI Security Institute (2026g). Our evaluation of Claude Mythos Preview's cyber capabilities. UK AI Security Institute. non-peer-reviewed Cited in §E.6
  405. UK AI Security Institute (2026h). Our evaluation of OpenAI's GPT-5.5 cyber capabilities. UK AI Security Institute. non-peer-reviewed Cited in §E.6
  406. UK AI Security Institute (2026i). Towards technical best practice for frontier AI development. UK AI Security Institute. non-peer-reviewed Cited in §E.6
  407. UK AI Security Institute (2026j). UK AISI / CAISI Preliminary Assessment of Kimi K3's Cyber Capabilities. UK AI Security Institute. non-peer-reviewed Cited in §E.6
  408. UK AI Security Institute (2026k). What can sandboxed AI agents learn about their evaluation environments? UK AI Security Institute. non-peer-reviewed Cited in §E.6
  409. US Center for AI Standards and Innovation (2026a). CAISI Assessment of Z.ai's GLM-5.2. US Center for AI Standards and Innovation. non-peer-reviewed Cited in §E.6
  410. US Center for AI Standards and Innovation (2026b). CAISI Evaluation of DeepSeek V4 Pro. US Center for AI Standards and Innovation. non-peer-reviewed Cited in §E.6
  411. US Center for AI Standards and Innovation (2026c). Towards Best Practices for Automated Benchmark Evaluations. US Center for AI Standards and Innovation. non-peer-reviewed Cited in §E.6
  412. Voudouris, K., Roberts-Gaal, X., Buhl, M., Irving, G., and Summerfield, C. (2026). AI alignment is a human problem. UK AI Security Institute. non-peer-reviewed Cited in §E.6
  413. Wagner, M., Roger, F., Cunningham, H., Gasteiger, J., Benton, J., and Mikulik, V. (2025). Training fails to elicit subtle reasoning in current language models. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  414. Wallace, E., Choquette-Choo, C. A., Kandpal, N., Toyer, S., Hunn, D., Lin, S., … Chen, K. (2026b). GPT-Red: Automated Red Teaming via Self-Play at Scale. arXiv. preprint Cited in §E.5
  415. Wang, A., Mahajan, P., Africa, D. D., Souly, A., Taylor, J., and Kirk, R. (2026f). Prefill Awareness in Large Language Models. arXiv. preprint Cited in §E.6
  416. Ward, J., Hu, S., Wong, A., Hu, N., Woodworth, B., Turner, A. M., and Cloud, A. (2026). Gradient Routing Localizes and Removes Unintended Behaviors in RL. arXiv. preprint Cited in §E.3
  417. Wichers, N., Ebtekar, A., Azarbal, A., Gillioz, V., Ye, C., Ryd, E., … Marks, S. (2025). Inoculation Prompting: Instructing LLMs to misbehave at train-time improves test-time alignment. arXiv. preprint Cited in §E.3
  418. Williams, M., Raymond, C., and Carroll, M. (2025). Sidestepping Evaluation Awareness and Anticipating Misalignment with Production Evaluations. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  419. Williams, M., Guo, P., Oks, D., Bemis, D., Staley, K., Wong, W., … Kivlichan, I. (2026a). How we detect frontier model misalignment. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  420. Williams, M., Sheahan, H., Raymond, C., Korbak, T., Pan, D., Yang, P., … Carroll, M. (2026b). Predicting LLM Safety Before Release by Simulating Deployment. arXiv. preprint Cited in §E.6
  421. Witten, Z., and Benton, J. (2025). Won't vs. Can't: Sandbagging-like Behavior from Claude Models. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
  422. Xiao, F., and Phuong, M. (2026). Generalization Hacking: Models Can Game Reinforcement Learning by Preventing Behavioral Generalization. arXiv. preprint Cited in §E.6
  423. Xu, X., Nitishinskaya, J., Schoen, B., Mossing, D., and la Tour, T. D. (2026). Studying metagaming latents in language models. OpenAI Alignment Research Blog. non-peer-reviewed Cited in §E.5
  424. Yona, I., Shumailov, I., Hayes, J., and Carlini, N. (2024). Stealing User Prompts from Mixture of Experts. arXiv. preprint Cited in §E.6
  425. Yueh-Han, C., McCarthy, R., Lee, B. W., He, H., Carroll, M., and Korbak, T. (2026). Reasoning Models Struggle to Control their Chains of Thought. Proceedings of the 43rd International Conference on Machine Learning (PMLR 306). Cited in §E.5
  426. Zhang, J., Sleight, H., Peng, A., Schulman, J., and Durmus, E. (2025c). Stress-Testing Model Specs Reveals Character Differences among Language Models. arXiv. preprint Cited in §E.3
  427. Zhang, J., Sleight, H., and Benton, J. (2026a). A3: An Automated Alignment Agent for Safety Finetuning. Anthropic Alignment Science Blog. non-peer-reviewed Cited in §E.3
AI Alignment: From the Optimization Problem to the Institutions Around It · © 2026 Changkun Ou · CC BY-NC-SA 4.0
Bibliography · PV/UV /