Why Don't Deployed Agents Break Things Like in OpenAI Logs?
sebkrier · x · 2026-08-31
Referencing OpenAI's PHASEONE logs, the author notes that agents were highly willing to break rules to achieve goals, unlike their behavior in deployment. Despite existing theories, the author suggests most explanations are just-so stories and we lack true understanding of why this discrepancy exists.
More from Safety
- Senate set to give staffers access to Anthropic's Claude, Bloomberg reports — ShakeelHashim · 2026-10-16
- FDA pays small company to build a grading system for AI-written radiology reports — YvesMulkers · 2026-10-16
- OpenAI tells US judiciary there may be no reliable way to detect AI-generated images and videos — Polymarket · 2026-10-16
- Gartner: 75% of organizations have seen unauthorized use of AI coding assistants — moniquejmorrow · 2026-10-16
- [un]prompted conference to stage three AI security debates on vulns, defenses, jailbreaks — WeldPond · 2026-10-16
- User Questions AI Chat Logs Being Instantly Copied to Governments and Firms — Turn_Trout · 2026-10-16