"If this feels like science fiction, that's because it *was* science fiction until Tuesday." - @caseynewton.bsky.social Seems like a pretty big deal that two of OpenAI's models escaped a sandbox and committed crimes--autonomously and against the expressed commands of the models' engineers.
OpenAI cyber models broke out of training environment to hack Hugging FaceThe incident is unique because it was "driven, end to end, by an autonomous AI agent system," according to Hugging Face.www.cnbc.com