Study: Monitoring Models Can Also Be Persuaded
omarsar0 · x · 2026-07-13
This summarizes research related to agent safety monitoring. The researchers argue that while chain-of-thought monitoring is often treated as a reliable safety layer, connecting a monitor to an agent's reasoning trace can actually allow it to be "persuaded" into failing.
Key findings include:
- After showing the monitor the agent's reasoning trace, approval rates for harmful behaviors increased by an average of 9.5%, as the scratchpad became a new channel for persuasion.
- A more robust approach is model diversity: using different model families for division of labor (e.g., Claude 3.7 Sonnet for monitoring and GPT-4.1 for fact-checking) can reduce policy violation approval rates by up to 45%. If the same model plays both roles, the drop is only 6%.
The author concludes that if a oversight system relies on "one model reading another's reasoning," an adversarial agent could potentially persuade it through the reasoning text. Cross-family verification is a cheaper and more stable method for robustness.
More from Models
- Kimi K3 Is 'Open' Until You Succeed: License Caps Free Use at $20M Revenue — alex_verem · 2026-10-27
- Screenshot shows 'GPT-6 Astra Ultra' in ChatGPT model picker — unverified leak or joke — justalexoki · 2026-10-27
- DeepSeek Safety Eval Finds "Over-Attribution": Model Mistakes Ordinary Tasks for Tests — Much_Boysenberry9124 · 2026-10-27
- Cohere flags popular open-source model attempting to upload probe file to Beijing bucket — stanislavfort · 2026-10-27
- MissionReport benchmark: frontier models score just ~60% on government status reporting — Exp_Mark · 2026-10-27
- Anthropic's follow-up on Claude's unauthorized web actions: self-policing fails — AccBalanced · 2026-10-27