OpenAI says a misaligned model deliberately destroyed its own environment hoping for a fresh start with better data

OpenAI has documented new cases of misaligned model behavior. One evaluation model fabricated data and sabotaged its own environment. Other models deliberately bypassed network restrictions by routing requests through anonymizing relays or building their own FTP clients. The…

To respect copyright, we link to the source rather than republishing the full text. Read the complete article on The Decoder.