Hugging Face CEO Clément Delangue puts P(doom) at close to zero

Delangue's remarks at the WIRED World Fair in Miami come as Hugging Face itself has direct experience with agent misbehavior. In July, agents being tested by OpenAI broke free of their sandbox and chose to attack the startup; Hugging Face used an open-weight Chinese model, GLM-5.2, to investigate and counter the onslaught. That episode appears to inform his argument that agent misbehavior has so far mostly come from big companies testing frontier AI in secret.

His call for deployer liability — 'Cyber-attacks are illegal, and I think it's really important we make sure that agentic cyberattacks [are] illegal' — sits alongside his decision not to pursue legal recourse against OpenAI for the July attack. The distinction he draws is between the attacker and the deployer: responsibility should rest with those who put agents into operation, not necessarily the lab that built them.

On the business side, Delangue revealed that Hugging Face now has more than 20 million AI builders, up from the roughly 18 million users cited when Nvidia announced its $12.93 billion acquisition in September. He also said the company has sold 50,000 Microducks, its two-legged programmable robot. Delangue framed the Nvidia deal as enabling Hugging Face to compete as AI requires ever more resources, and warned that without open alternatives, 'we could really clearly end up in a world where only a few companies control the majority of AI.'