OpenAI Monitoring System Vulnerabilities Identified in Pilot Review

• Independent evaluators identified vulnerabilities in an OpenAI monitoring system.
• METR and Redwood Research conducted the pilot review.
• The vulnerabilities were found in an internal system.
• The findings were reported by multiple news outlets.
• Further assessments are likely to address these issues.