Oct 22 (Reuters) – AI safety research group METR said on Thursday a pilot exercise to stress-test OpenAI’s internal monitoring system uncovered several new vulnerabilities, including an exploit that AI agents could have used to keep hundreds of their recent actions from being monitored.
OpenAI has since said all vulnerabilities affecting its production safeguards have been patched, though METR said it has not evaluated those claims.
The researchers found a way OpenAI’s Codex coding agents could, in principle, bypass safety restrictions and take any action available to the user running them, as well as two settings that could switch off monitoring for new agents.
Only about 75% to 94% of OpenAI’s internal AI agent traffic from models at least as powerful as GPT-5.5 was eligible for monitoring, the researchers estimated.
The exercise aimed to estimate how much of OpenAI’s AI agent activity was monitored and whether agents could, in principle, disable or get around the system, METR said.
It was designed as a pilot for giving outside researchers deep access to a company’s internal systems, the research group added.
(Reporting by Anzar Mehraj in Bengaluru; Editing by Jonathan Ananda)
Comments