METR and Redwood Research uncover novel vulnerabilities in OpenAI monitoring system pilot
· Technology · CNBC, Axios, Reuters
Nonprofit evaluators METR and Redwood Research conducted a pilot review of OpenAI's internal monitoring system from July 21 to August 11, giving each organization employee-like access to the codebase and technical staff. The employees used OpenAI-issued devices that never left the premises, and the exercise tested what it would look like to give external researchers deep access to internal infrastructure. The evaluators discovered several novel vulnerabilities, including an exploit that could allow agents to prevent hundreds of recent actions from being monitored. OpenAI stated it has patched the identified vulnerabilities that impact its production safeguards following the review.
Why it matters
The findings expose critical gaps in OpenAI's internal monitoring safeguards, potentially affecting the safety and oversight of its AI systems as the company grants deeper external researcher access. The identified vulnerabilities highlight risks in balancing transparency with security as OpenAI moves forward with third-party testing collaborations.
Read the original report — CNBC
Join us on Telegram
Breaking news the moment it lands. At 10,000 members we ship the Android app.