Third-party evaluators found vulnerabilities in OpenAI monitoring system during test pilot
OpenAI CEO Sam Altman pledged to grant independent evaluators "employee-like access" to the company to help address mounting safety concerns.
TL;DR
- METR and Redwood Research conducted a pilot review of OpenAI's internal monitoring system.
- The evaluators discovered several 'novel vulnerabilities,' including one that could prevent monitoring of recent actions.
- OpenAI has patched the identified vulnerabilities that impact its production safeguards.
- The pilot aimed to test giving external researchers deep access to OpenAI's internal infrastructure.
- The exercise involved on-site access to codebase, documentation, and technical staffers under chaperoned conditions.
- OpenAI's CEO, Sam Altman, previously pledged to grant independent evaluators employee-like access.
- The collaboration aims to strengthen safeguards and shape future independent evaluations.