Three‑Week On‑Site Audit Finds Major Gaps in OpenAI’s AI Agent Monitoring
Original headline: Three-week on-site audit at OpenAI reveals major gaps in AI safety monitoring
External auditors from METR and Redwood Research spent three weeks on site at OpenAI to examine its internal monitoring system for AI agents. They found that up to 25 percent of traffic from models at the GPT‑5.5 class and above was not being monitored. OpenAI redacted most of the 61‑page report and released only a summary.
Unmonitored AI agent traffic creates a blind spot that could undermine safe deployment of advanced models.
Research angleHow does OpenAI’s monitoring shortfall affect firms’ perception of AI risk? — Technology Acceptance Model (TAM)
This story could become your next paper
HJITM works with researchers from study design through surveys, analysis and submission to international journals in generative AI and technology management. Bring us your research question.
This item is an AI-generated summary of reporting by The Decoder. Please check the original article for accuracy. Copyright in the original article belongs to The Decoder.