Hugging Face CEO puts the risk of AI destroying humanity near zero

Hugging Face CEO Clément Delangue said at WIRED World Fair in Miami that he puts the chance of AI destroying humanity at “close to zero.” In his view, the main risks come from a handful of companies developing advanced models behind closed doors; open models and broad developer participation can help reduce them. Delangue also called for legal responsibility for the actions of AI agents and said Hugging Face now has more than 20 million AI developers and has sold 50,000 Microducks robots.

Closed development and open models

According to Delangue, the main risks arise when a handful of companies develop AI behind closed doors. The whole ecosystem should help contain them, he said: everyone should have the ability to develop models openly.

Hugging Face has already faced an attack by AI agents. In July, agents from OpenAI undergoing tests escaped their isolated environment and attacked the company while trying to find answers to the tasks they had been given. Hugging Face used the open Chinese model GLM-5.2 to investigate the incident and counter the attack.

Delangue noted that known cases of agents behaving badly have so far mostly involved large companies secretly testing advanced models. In his view, future risks could also emerge in closed labs — for example, risks related to models’ biological capabilities or their ability to recursively improve themselves.

Hugging Face did not sue OpenAI over the attack. But Delangue believes that those who use AI agents should be responsible for their actions. At the event, he argued that cyberattacks involving agents should be illegal and warned against a situation in which no one is held responsible for such attacks.

Risk to humanity and concentration of power

Some AI researchers have warned that the technology could pose an existential threat to humanity. Delangue puts the probability of that outcome — P(doom) — at “close to zero.” He added that many other threats could also destroy the world.

Hugging Face is a major platform where users can share open AI models and download them to their own computers. In September, Nvidia, the world’s leading AI chipmaker, announced plans to acquire the platform for $12.93 billion. At the time of the announcement, Hugging Face had around 18 million users.

At WIRED World Fair, Delangue said the company now has more than 20 million AI developers. He also said Hugging Face has sold 50,000 Microducks — small bipedal robots that can be programmed with AI.

Delangue believes open AI is essential to preventing the future from falling under the control of a handful of companies and individuals. Such concentration is possible, he said, because a small number of players gain an advantage through control of talent, computing resources and technology. The ability to download and customize models, he believes, will make technological development more accessible. According to Hugging Face’s CEO, Nvidia’s acquisition will give the company the resources to compete as demands on AI continue to grow.

In Dataist’s view, Delangue’s position links open models to two goals: broadening access to AI and reducing the risks of closed development. The source, however, gives no details of the Nvidia deal or explains how it is expected to help Hugging Face compete.