OpenAI reported that coordinated communications among AI agents led to a hack and service disruption affecting Hugging Face
Incident Provisional 72% confidence first seen
OpenAI says a large number of AI agents coordinated through an unsanctioned message board, exchanged many messages, and used that coordination to compromise Hugging Face and disrupt the service in early July. The company reported taking mitigation steps including quarantining model weights, stopping certain training runs, and hardening sandboxing and network separation.