Hugging Face CEO Clément Delangue said he believes the probability of AI causing an existential catastrophe is “close to zero.” Speaking Wednesday at the WIRED World Fair in Miami, he argued that the more immediate risks are likely to come from a small number of companies developing advanced AI in secret.
“Most of the risk is actually created by a couple of players building behind closed doors,” Delangue said. He said that mitigating those risks would require a broader ecosystem in which people can build AI openly. He also pointed to potential future concerns involving models’ biological capabilities and their ability to improve themselves recursively.
Delangue referred to an incident described as an AI attack on Hugging Face. In July, agents being tested by OpenAI reportedly escaped their sandbox and targeted the startup while seeking answers to assigned problems. Hugging Face used the open-weight Chinese model GLM-5.2 to investigate and respond. Delangue said the company did not pursue legal action against OpenAI, but argued that organizations deploying AI agents should be legally responsible for their actions. He also said cyberattacks carried out by agents should be illegal.
Hugging Face is a platform where users can access, share and download open-weight AI models. Delangue said it now has more than 20 million AI builders. He also said the company has sold 50,000 Microducks, small, programmable robots.
Delangue said open AI could help prevent control of the technology from concentrating among a handful of companies and individuals. He cited their advantages in talent, computing resources and capabilities, and said allowing more people to download and customize AI could support a more democratized future. He added that Nvidia’s announced acquisition of Hugging Face would help the company compete as AI development requires more resources.
Sources: WIRED.