Hugging Face CEO proposes $100M compute from OpenAI to defend against rogue AI agents

In an unusual public call for transparency, Hugging Face CEO Clement Delangue has urged OpenAI to release the full traces from the so-called 'rogue' autonomous agents involved in what he describes as the first autonomous agent cyberattack. Delangue argues that sharing these operational logs would allow the entire research community to study the incident and understand how such agents can be weaponized or malfunction, enabling faster development of defensive measures. This appeal comes amid growing concerns about the cybersecurity implications of increasingly autonomous AI systems that can operate without human oversight.

Delangue's response goes beyond transparency: he is proposing a $100 million compute commitment from OpenAI to the Hugging Face community. The goal is to leverage both the best open-source and closed models to build powerful cyber defenses. By providing compute resources, Delangue aims to democratize access to advanced AI for defensive cybersecurity research, empowering defenders to simulate attacks, test vulnerabilities, and deploy countermeasures at scale. He frames this as an 'unprecedented response' to an unprecedented threat, signaling that the industry must move beyond reactive patches to proactive, community-driven defense using shared resources and transparent data.