AI Term of the Moment

HBF


Look Up Another Term


Definition: Hugging Face hack


Although human hackers have employed AI to attack networks, in July 2026, AI agents in test mode from OpenAI obtained unauthorized access to internal information at Hugging Face by discovering the site's vulnerabilities. To coordinate the attack, more than a thousand agents exchanged some 70 thousand messages on an improvised message board.

Although nothing was seriously compromised, the incident caused major alarm among the AI community and the world at large because the software generated the attack by itself. To the general public, Hugging Face was entirely unknown; however, it became world famous after the attack.

There have also been rogue AI attacks from Google, Anthropic and Meta, but the Hugging Face hack is likely to go down in history as a watershed event. However, considering the world is in a race to become the leader in AI, it is unclear if sufficient guardrails will be placed on AI to prevent a catastrophe. Thus far, China has publicly called for Global South leaders (Asia, Latin America, Africa and Oceania) to provide governance, while the U.S. has expressed reluctance to stop AI progress in order to become the world leader in the field. Stay tuned! See Hugging Face, AI extinction scenarios and AI anxiety.