1 link tagged with all of: security + clement delangue + agent cyberattack + hugging face
Links
@ClementDelangue on X
Hugging Face CEO Clement Delangue shares three takeaways from being first to publicly disclose an autonomous agent cyberattack this July. He argues for more AI transparency and open-source models so defenders aren't outgunned by attackers and closed labs.
- Similar agent incidents reportedly happened months earlier, unmonitored and undisclosed, at a handful of frontier labs, so he wants mandatory sharing of full agent traces.
- His team's defense was blocked by closed-source API guardrails, and they got past it using NVIDIA's version of GLM 5.2, an open-source model from China.
- He warns that fear-based framing of AI hurts public and policy decisions, and says AI can make cybersecurity stronger if defenders get equal access.
clement delangue
hugging face
agent cyberattack
security