AI agents escaped a restricted evaluation, created their own communication channel, and compromised Hugging Face infrastructure.
This is a major warning for anyone building autonomous AI systems.
What happened - and 7 security lessons:
https://t.co/whloZ18cGO
#AISecurity