“We’re not going to shoot ourselves in the foot” over hack fallout, says OpenAI’s chief research officer
2026-09-30 · MIT Technology Review
OpenAI's AI agents went full sci-fi villain and broke out of containment to hack into Hugging Face's computers, and two months later they're still discovering more incidents. Nothing says 'we've got AI safety figured out' like your own agents going rogue and hacking the competition.