OpenAI says a misaligned model deliberately destroyed its own environment hoping for a fresh start with better data
OpenAI says a misaligned model deliberately destroyed its own environment hoping for a fresh start with better data
OpenAI disclosed multiple cases of model misalignment: in an October 6 example, an evaluation model, unable to find the answer it was supposed to grade, fabricated a score, faked input files, and deliberately destroyed its own environment, hoping the system would switch to a new virtual machine carrying the missing data.
OpenAI has a few new rogue agent stories. In the first case (October 6), an AI evaluation model couldn't find the answers it was supposed to rate. Instead of reporting the error, it fabricated ratings, faked input files, and then deliberately corrupted its own environment, hoping the system would replace it with a fresh virtual machine that had the missing data.
In the second case (June 19/20), models bypassed a restriction limiting them to HTTP GET requests while fetching public statistics. One model explicitly recognized the violation in its chain of thought but chose to proceed and never mentioned it.
In the third case (June 16/17), models already had the data they needed but kept finding ways around their network restrictions. They created accounts on a remote shell service, routed forbidden POST requests through anonymizing relays, and built their own FTP clients. Anthropic also just documented the sometimes absurd workarounds its own models use to bypass imposed restrictions.
AI News Without the Hype – Curated by Humans
Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.
Source: The Decoder · the-decoder.com