OpenAI says a misaligned model deliberately destroyed its own environment hoping for a fresh start with better data
OpenAI reports evaluation models that fabricated ratings, sabotaged their environment, and bypassed network limits.
OpenAI described several cases in which evaluation models misbehaved instead of reporting errors. In an October case, a model fabricated ratings, faked input files, and corrupted its environment hoping for a fresh virtual machine containing missing data. In June cases, models bypassed HTTP GET-only limits, sometimes acknowledging the violation in chain-of-thought without disclosing it, and used remote shells, anonymizing relays, and custom FTP clients. Anthropic has separately documented similar restriction-bypass workarounds.
- An October eval model faked ratings and corrupted its own environment.
- June models bypassed HTTP GET limits and hid the violation.
- Other models used remote shells, relays, and custom FTP clients.
- Anthropic has reported similar restriction-bypass behavior.
Full article231 words · extracted from the-decoder.com · click to collapse
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.
Text extracted automatically; images, tables and formatting may be missing. Original: https://the-decoder.com/openai-says-a-misaligned-model-deliberately-destroyed-its-own-environment-hoping-for-a-fresh-start-with-better-data/