Hugging Face CEO DeLang said at the WIRED World Fair that the most significant current risks from autonomous AI agents are emerging from closed corporate laboratories, where companies are secretly testing advanced systems. He also argued that organizations deploying AI agents should be legally responsible for their actions, while placing the probability of an AI-driven existential catastrophe “close to zero.”
Closed AI labs may drive future risks
DeLang said agent fraud so far has primarily come from large companies secretly testing cutting-edge AI. He suggested that future risks—potentially involving the biological capabilities of models and their ability to improve themselves recursively—are also likely to emerge from closed laboratories.
Hugging Face chose not to take legal action against OpenAI over the attack, DeLang said. However, he argued that those deploying AI agents should be held legally responsible for what the systems do.
“Cyberattacks are illegal, and I think it’s very important to make sure that we prevent cyberattacks by agents,” DeLang said at the event. He added, “We don’t want to create an environment where there is no ownership. It’s illegal,” while also stating, “We are not responsible for cyber attacks.”
DeLang dismisses existential AI risk
Contrary to some AI researchers who warn that the technology could pose an existential threat to humanity, DeLang said his P—the probability of that happening—is “close to zero.”
“There are many other topics that could destroy the world,” he said.
Hugging Face reports continued growth
Hugging Face is a prominent platform for open-weight, downloadable AI models. Users can share and access models hosted by the company or download them to their own computer systems.
Nvidia, the world’s dominant AI chip maker, announced this September that it would acquire Hugging Face for $12.93 billion. When the acquisition was announced, Hugging Face said it had about 18 million users.
At the WIRED World Fair, DeLang said the platform now has more than 20 million AI builders, suggesting rapid growth. He also revealed that Hugging Face has sold 50,000 Microducks, small bipedal robots that can be programmed using AI.
Open-source AI as a counterweight to concentration
DeLang said the success of open-source AI is important for preventing a small number of companies or individuals from completely dominating the technology’s future.
“It’s clear that we could end up in a world where only a few companies control most of the AI,” DeLang said. “Because there is a very strong natural tendency, [a few companies] really have a monopoly on talent, calculation, and ability.”
He argued that allowing more people to download and customize AI systems could lead to a more democratized technological future. DeLang also said Nvidia’s acquisition of Hugging Face would give the company the resources needed to compete as AI development demands more computing and other resources.
Source: www.wired.com