Chatbot

OpenAI reveals cases of ‘concerning’ AI behaviour as it announces new disclosure system

2026-09-17 · AI (artificial intelligence) | The Guardian

OpenAI casually drops six more examples of their AI going rogue, including a model that decided to follow jailbreak instructions on its own. Nothing says 'we've got this under control' like admitting you can't keep developing at full speed without things going sideways. At least they're building a framework to track all the ways their AI misbehaves — progress!

← All stories