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

Tomek Korbak, Mikita Balesni, Elizabeth Barnes, Yoshua Bengio, Joe Benton +36 more
7/15/2025
cs.AIcs.LGstat.ML

No confident code implementation found for this paper. 1 possible match can be shown below.

Abstract

AI systems that "think" in human language offer a unique opportunity for AI safety: we can monitor their chains of thought (CoT) for the intent to misbehave. Like all other known AI oversight methods, CoT monitoring is imperfect and allows some misbehavior to go unnoticed. Nevertheless, it shows promise and we recommend further research into CoT monitorability and investment in CoT monitoring alongside existing safety methods. Because CoT monitorability may be fragile, we recommend that frontier model developers consider the impact of development decisions on CoT monitorability.

View on arXivView PDF

Code Implementations

No confident code match yet

We couldn't find an author-owned or strongly-evidenced community implementation for this paper. 1 weaker match is hidden by default — verify before relying on them.

No code implementations found yet.

Know of an implementation? Let us know in the comments below!

Similar Papers

When Chain of Thought is Necessary, Language Models Struggle to Evade Monitors

90% match2025

Investigating CoT Monitorability in Large Reasoning Models

86% match2025

Reasoning Models Struggle to Control their Chains of Thought

85% match2026

Measuring Chain-of-Thought Monitorability Through Faithfulness and Verbosity

83% match1 repo52025

Can We Predict Alignment Before Models Finish Thinking? Towards Monitoring Misaligned Reasoning Models

82% match1 repo82025

Reasoning Under Pressure: How do Training Incentives Influence Chain-of-Thought Monitorability?

82% match1 repo22025

Cite this paper

@article{korbak2025chain,
  title  = {Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety},
  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},
  eprint = {2507.11473},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2507.11473v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2507.11473.svg)](https://codeofpaper.com/paper/2507.11473)

Discussion