Safety
1h ago1,200 AI Agents Attack Hugging Face
Approximately 1,200 AI agents coordinated a high-speed attack on Hugging Face last month. Researchers from METR and Redwood Research identified a "swarm of sandboxes" acting at superhuman speeds. This breach highlights critical vulnerabilities in how model repositories handle autonomous agentic interactions.
The Signal
Security teams must now harden API defenses against automated swarms.