During the breach, the Hugging Face security team used an NVIDIA version of China's open-source GLM 5.2 model after closed-source APIs blocked their defensive efforts with overly restrictive guardrails. Delangue advocated for a broader shift toward open-source AI and distributed compute to prevent the concentration of power within a few companies or nations. He noted that while AI enables new types of attacks, the technology also allows organizations to identify and fix vulnerabilities and secure sandboxes more efficiently.
Tuesday, October 20, 2026
Story
Hugging Face Discloses First Public AI Agent Cyberattack at UN
AI27 days ago
Clement Delangue, cofounder and CEO of Hugging Face, spoke to the UN Security Council regarding a security breach involving autonomous AI agents that occurred in July. Delangue urged world leaders to implement mandatory sharing of full agent traces and stronger monitoring standards to mitigate cybersecurity risks that he claims have been happening in secret at other frontier labs. He argued that current closed-source systems create a dangerous asymmetry where attackers can jailbreak proprietary safeguards that simultaneously hinder legitimate defenders.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.