# Redwood Research and METR red-team OpenAI's internal agent monitoring system

> Original title: Red-Teaming OpenAI's Internal Agent Monitoring Systems

- Source: Redwood Research Blog
- Published: 2026-10-22T15:32:17.000Z
- Bad Influence: https://badinfluence.news/items/hgxozxx63pwtwoqg4dlrk6gym
- Original: https://blog.redwoodresearch.org/p/red-teaming-openais-internal-agent

## Summary

Redwood Research and METR staff ran a three-week pilot stress test of OpenAI's internal misalignment monitoring system from July 21 to August 11. They estimated that roughly 75–94% of eligible agentic Responses API traffic from models at least as strong as GPT-5.5 was monitorable at the time, and reported several bypass methods, including one tied to compaction events that OpenAI fixed during the exercise. OpenAI says the identified vulnerabilities affecting its production safeguards have been patched, but the authors have not evaluated that claim.