AIThe Decoder1h ago
Three-week on-site audit at OpenAI reveals major gaps in AI safety
Three-week on-site audit at OpenAI reveals major gaps in AI safety monitoring
TL;DROpenAI's safety monitoring misses up to 25% of traffic from its most advanced AI models.
Why it matters: Unmonitored AI agents at frontier capability levels pose unknown risks that regulators and users can't assess.
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.…
Read full articleSource: The Decoder · Opens in new tab