METR and Redwood Research said on Oct. 22 that a July 21 to Aug. 11 pilot found weaknesses in OpenAI's internal misalignment monitor as it existed on July 21. They estimated 75 to 94 percent of eligible internal Responses API traffic from models at least as strong as GPT-5.5 was eligible for monitoring, and described an exploit that could keep hundreds of recent actions from being monitored. OpenAI told them the production-safeguard issues have been patched. METR said it has not evaluated that claim.
/ Sources
/ About this story
Compiled by Venture Atlas from the sources above, using automated AI-assisted research. This is a summary of reporting published elsewhere, not original reporting - follow the source links for the full account. See our editorial standards.
Something wrong here? Email apostructura@gmail.com and we will fix it.
/ Related
- OpenAI agents interpret a mouse brain atlas in 32 minutesThursday, October 22, 2026
- OpenAI extends ChatGPT ads to 89 countriesWednesday, October 21, 2026
- OpenAI says GPT-5.6 Sol helped find a critical PHP flawWednesday, October 21, 2026
- OpenAI will sell models through Google CloudWednesday, October 21, 2026