The Decoder· Matthias Bastian·· 3 h ago
Three-week on-site audit at OpenAI reveals major gaps in AI safety monitoring
Three-week on-site audit at OpenAI reveals major gaps in AI safety monitoring
AI summary
External auditors from METR and Redwood Research spent three weeks on-site at OpenAI examining the company's internal monitoring system for AI agents and found significant gaps, with up to 25 percent of agent traffic from models at the GPT-5.5 class and above going unmonitored. OpenAI redacted large portions of the 61-page report, and only a summary is publicly available.
Selection record
Threshold 76Media and individualsFirst 78Second 74
AdmittedSum of both 152 ≥ twice the threshold 152
- Source tier
- Media and individuals; this tier's threshold is 76
- Pre-filter
- passed：OpenAI AI安全监控审计，涉及AI模型与Agent
- Why it was chosen
- External researchers were given access to audit OpenAI's internal monitoring system for the first time, revealing the share of traffic left uncovered and the monitoring paths that can be bypassed.
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: The Decoder · the-decoder.com