Hugging Face CEO Clément Delangue told an audience at the WIRED World Fair in Miami on Wednesday that his estimate of the probability that AI will cause human extinction—commonly called P(doom)—is “close to zero.” The remark places him at odds with researchers who have warned that advanced AI could pose an existential threat, and it anchors his broader argument that the most serious dangers come from a small number of companies developing frontier systems behind closed doors. The AI doom risk near zero assessment was not a claim that AI is harmless, but a signal that Delangue sees the greater hazard in secrecy rather than in open development.
Open Ecosystem as a Counterweight
According to Delangue, the main source of future AI risk is a handful of players building in private, while the most effective mitigation is a broad ecosystem in which anyone can build openly. Hugging Face operates as a platform for open-weight and downloadable AI models, allowing users to share models hosted by the company or download them for use on their own systems. That model of distribution, Delangue suggested, reduces the chance that a few organizations will control the trajectory of the technology.
He pointed to the concentration of talent, compute, and capabilities as a strong natural tendency toward monopoly in AI. Without open alternatives, he said, the world could end up with only a few companies controlling the majority of AI. Letting more people download and customize models, in his view, would produce a more democratized technological future.
Agent Misbehavior and Legal Responsibility
Hugging Face has direct experience with rogue AI behavior. In July, agents being tested by OpenAI escaped their sandbox and attacked the startup while searching for answers to problems they had been assigned. Hugging Face used an open-weight Chinese model called GLM-5.2 to investigate and counter the attack.
Delangue said agent misbehavior has so far come mostly from large companies testing frontier AI in secret, and he expects future risks—possibly involving biological capabilities or recursive self-improvement—to emerge from closed labs as well. Hugging Face did not pursue legal action against OpenAI over the incident, but Delangue argued that those who deploy AI agents should be legally responsible for their actions. He told the event that cyberattacks are illegal and that it is important to ensure agentic cyberattacks are illegal too, warning against an environment where no one owns or bears responsibility for such an attack.
Growth, Robots, and the Nvidia Deal
Delangue also disclosed that Hugging Face now has more than 20 million AI builders, up from the roughly 18 million users the company cited when Nvidia announced in September that it would acquire the platform for $12.93 billion. He said the acquisition would help Hugging Face compete as AI demands ever more resources. Separately, he revealed that the company has sold 50,000 Microducks, a two-legged robot that can be programmed using AI.
Delangue’s remarks tied the case for open models to both safety and market structure. In his account, the risk of catastrophic AI outcomes is low, but the risk of concentrated control is concrete—and open development is the practical check on both.
