How to build fast, efficient monitors for AI models using probes - Goodfire

An introduction to using probes for monitoring AI models: what probes are, when to use them, and how to build them.

AI agents can now take thousands of actions faster than a human can review them. The recent Hugging Face security incident starkly demonstrated why this matters: without monitors, agents can take drastically undesirable actions at scale without anyone noticing. If you build or serve AI models, you need monitors to know when models are trying to take unintended actions, and to prevent them when they do. One of the best tools in the toolbox for building these monitors is activation probes. This post is an introduction to what probes are, when to use them, and how to build them. What are…

saved by

• Ishaan Panigrahi

related reading

• The fragile foundations of CoT monitoring | Christopher Pottsweb.stanford.edu
• [2512.11949] Neural Chameleons: Language Models Can Learn to Hide Their Thoughts from Unseen Activation Monitorsarxiv.org
• Should We Train Against (CoT) Monitors? — LessWronglesswrong.com
• [2506.10805] Detecting High-Stakes Interactions with Activation Probesarxiv.org
• Early work on monitorability evaluations - METRmetr.org
• [2507.11473] Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safetyarxiv.org
• [2506.10805] Detecting High-Stakes Interactions with Activation Probesarxiv.org
• Inside a Neural Chameleonjacksonmowattgok.com
• Coup probes: Catching catastrophes with probes trained off-policy — LessWronglesswrong.com
• Scaling Activation Oracles to Trillion-Parameter Modelstransluce.org
• [2601.11516] Building Production-Ready Probes For Geminiarxiv.org
• Proposal for tracking the effects of architecture on monitorability — Redwood Researchredwoodresearch.org