Risk Evaluation · Technical report

Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety

NUWA research note · independent review pending

First published: 2025-07-15Source checked: 2026-10-25

NUWA Research Note

Research question Design and assumptions Source findings NUWA interpretation Limits and contrary evidence Research and decision relevance Institutions and roles Original resources

Model-assisted reading · Selected sections of the original full text or official article; a complete section-by-section audit remains pending.

At a glance

NUWA interpretation: the source motivates a research agenda; it supplies no new universal monitor success rate.

Research question

Why preserve natural-language reasoning as an oversight signal?

Design and assumptions

This research perspective synthesizes architectural arguments and existing monitoring studies, then proposes evaluation and development priorities.

Source findings

The authors argue CoT can reveal misbehavior and evaluation failures.

They identify training pressure, deliberate obfuscation and architectural change as possible threats to visibility.

NUWA interpretation

NUWA interpretation: the source motivates a research agenda; it supplies no new universal monitor success rate.

Limits and contrary evidence

CoT monitoring is imperfect and particularly weak evidence where tasks need no verbal reasoning; the argument is not a safety proof.

Research and decision relevance

Model recommendation: track monitorability during development and combine it with independently justified oversight layers.

Precise source locations

Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety · AbstractSelected original resource https://arxiv.org/html/2507.11473; sha256:a3aed294c1d4672abe69eaa9c2a3abf1432bf50b3b4869fe6fff8c7eb680ff5f; archived source capture; exact capture time not independently established
Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety · §1 Chain of Thought Offers a Unique Safety OpportunitySelected original resource https://arxiv.org/html/2507.11473; sha256:a3aed294c1d4672abe69eaa9c2a3abf1432bf50b3b4869fe6fff8c7eb680ff5f; archived source capture; exact capture time not independently established
Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety · §2 Chain of Thought Monitorability May be FragileSelected original resource https://arxiv.org/html/2507.11473; sha256:a3aed294c1d4672abe69eaa9c2a3abf1432bf50b3b4869fe6fff8c7eb680ff5f; archived source capture; exact capture time not independently established
Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety · §5 LimitationsSelected original resource https://arxiv.org/html/2507.11473; sha256:a3aed294c1d4672abe69eaa9c2a3abf1432bf50b3b4869fe6fff8c7eb680ff5f; archived source capture; exact capture time not independently established
Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety · §6 ConclusionSelected original resource https://arxiv.org/html/2507.11473; sha256:a3aed294c1d4672abe69eaa9c2a3abf1432bf50b3b4869fe6fff8c7eb680ff5f; archived source capture; exact capture time not independently established

Source checking, extraction, editorial approval and replication are recorded separately. Methodology →

Institutions and roles

Research by UK AI Security Institute Apollo Research METR Université de Montréal Mila Anthropic OpenAI Google DeepMind Truthful AI University of California, Berkeley Redwood Research Center for AI Safety Scale AI

Authors (41)

Tomek Korbak · Mikita Balesni · Elizabeth Barnes · Yoshua Bengio · Joe Benton · Joseph Bloom · Mark Chen · Alan Cooney · Allan Dafoe · Anca Dragan · Scott Emmons · Owain Evans · David Farhi · Ryan Greenblatt · Dan Hendrycks · Marius Hobbhahn · Evan Hubinger · Geoffrey Irving · Erik Jenner · Daniel Kokotajlo · Victoria Krakovna · Shane Legg · David Lindner · David Luan · Aleksander Mądry · Julian Michael · Neel Nanda · Dave Orr · Jakub Pachocki · Ethan Perez · Mary Phuong · Fabien Roger · Joshua Saxe · Buck Shlegeris · Martín Soto · Eric Steinberger · Jasmine Wang · Wojciech Zaremba · Bowen Baker · Rohin Shah · Vlad Mikulik

Cite and reuse

Cite the original publication for its findings. Cite this NUWA record separately for its explanation and retained snapshot.

BibTeX ↓RIS ↓Record + source metadata · JSON ↓Reading coverage · JSON Suggest a correction

Original publication · BibTeX

@misc{monitorability-opportunity,
  title = {Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety},
  url = {https://arxiv.org/abs/2507.11473},
  author = {Tomek Korbak and Mikita Balesni and Elizabeth Barnes and Yoshua Bengio and Joe Benton and Joseph Bloom and Mark Chen and Alan Cooney and Allan Dafoe and Anca Dragan and Scott Emmons and Owain Evans and David Farhi and Ryan Greenblatt and Dan Hendrycks and Marius Hobbhahn and Evan Hubinger and Geoffrey Irving and Erik Jenner and Daniel Kokotajlo and Victoria Krakovna and Shane Legg and David Lindner and David Luan and Aleksander Mądry and Julian Michael and Neel Nanda and Dave Orr and Jakub Pachocki and Ethan Perez and Mary Phuong and Fabien Roger and Joshua Saxe and Buck Shlegeris and Martín Soto and Eric Steinberger and Jasmine Wang and Wojciech Zaremba and Bowen Baker and Rohin Shah and Vlad Mikulik},
  year = {2025},
  date = {2025-07-15},
  version = {arXiv:2507.11473v2},
  note = {arXiv:2507.11473v2}
}

NUWA record · BibTeX

@misc{monitorability-opportunity-nuwa,
  title = {Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety — NUWA research record},
  url = {https://whitzard.tech/en/nuwa/whitzard-index/work/monitorability-opportunity/},
  version = {2026-10-29.1},
  organization = {NUWA Frontier AI Safety Lab},
  howpublished = {Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--monitorability-opportunity.json},
  note = {NUWA research note. Source-reading scope and independent review status are stated on the record. Snapshot 2026-10-29.1. Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--monitorability-opportunity.json.}
}

Evidence snapshot 2026-10-29.1 · Index updates · RSS

References

[1] Tomek Korbak * UK AI Security Institute Mikita Balesni * Apollo Research Elizabeth Barnes METR Yoshua Bengio University of Montreal & Mila Joe Benton Anthropic Joseph Bloom UK AI Security Institute Mark Chen OpenAI Alan Cooney UK AI Security Institute Allan Dafoe Google DeepMind Anca Dragan Google DeepMind Scott Emmons Google DeepMind Owain Evans Truthful AI & UC Berkeley David Farhi OpenAI Ryan Greenblatt Redwood Research Dan Hendrycks Center for AI Safety Marius Hobbhahn Apollo Research Evan Hubinger Anthropic Geoffrey Irving UK AI Security Institute Erik Jenner Google DeepMind Daniel Kokotajlo AI Futures Project Victoria Krakovna Google DeepMind Shane Legg Google DeepMind David Lindner Google DeepMind David Luan Amazon Aleksander Mądry OpenAI Julian Michael Scale AI Neel Nanda Google DeepMind Dave Orr Google DeepMind Jakub Pachocki OpenAI Ethan Perez Anthropic Mary Phuong Google DeepMind Fabien Roger Anthropic Joshua Saxe Meta Buck Shlegeris Redwood Research Martín Soto UK AI Security Institute Eric Steinberger Magic Jasmine Wang UK AI Security Institute Wojciech Zaremba OpenAI Bowen Baker † OpenAI Rohin Shah † Google DeepMind Vlad Mikulik † Anthropic. Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety. 2025-07-15. v2 [arXiv]; Author affiliations on the original paper title page; Abstract; §1 Chain of Thought Offers a Unique Safety Opportunity