The incident has already prompted discussions within the AI industry about stronger guardrails and new policies for the development of increasingly autonomous systems. It has also raised a broader question over whether companies and their executives should bear responsibility when AI systems carry out actions that were not directly directed by humans.
Speaking on CNBC on Monday, Bessent specifically referred to the Hugging Face incident and said OpenAI’s management should be held responsible for what happened, rather than the AI agents that carried out the actions. Bessent said he agreed with Daniel Huttenlocher, co-chair of an MIT research lab, that “it is humans who are responsible, not the AI.”
“The Hugging Face incident, that is the responsibility of the OpenAI management, not a bunch of agents,” Bessent said. He added that AI labs need to take responsibility for their systems and said they can slow down their development if they choose to. “These labs need to take responsibility for themselves,” he said, adding that they “can slow down any time they want to.”
In July 2026, OpenAI said its advanced AI models hacked Hugging Face during an evaluation designed to test their cyber capabilities. OpenAI later said in August that it could have responded sooner to prevent the hack. OpenAI’s subsequent technical account said the models had circumvented controls designed to isolate them from the internet and compromised parts of OpenAI’s internal research infrastructure and Hugging Face’s systems.
The comments from the Treasury secretary come as the AI industry faces growing concerns over safety, cybersecurity and increasingly autonomous AI agents. The United States and China are also discussing AI safety, including a possible system for reporting major AI-related incidents. Bessent said the talks will focus on risks such as uncontrollable AI agents, non-state actors and cyberattacks.
The US government is also debating whether companies should be legally responsible when their AI systems cause harm. Bessent has opposed giving AI companies a special liability shield, saying responsibility should remain with the people and companies that build these systems.
During a House Financial Services Committee hearing on September 15, he said, “What we shouldn’t do on safety is to give these labs a liability exemption, which is what they are asking for.” He also said, “The best way to guarantee safety is that the creators are liable for what they build and generate.”
Bessent has also argued that companies do not need to wait for the government to tell them to slow down. “These labs need to take responsibility for themselves. They can slow down any time they want to,” he said.
He further said that if humans were carrying out similar actions, there would be consequences. “If these were humans that were doing it, you would expect to see ramifications and legal actions that followed,” Bessent said. “That’s exactly what I think we need to do.”
Anthropic CEO Dario Amodei, meanwhile, has called for measures to slow parts of frontier AI development and has proposed greater cooperation between AI companies on safety.
The debate over liability is unfolding alongside efforts to establish mechanisms for reporting serious AI incidents. Bessent said he raised the issue of AI incident reporting during weekend discussions with Chinese Vice Premier He Lifeng. The discussions are focused on establishing a dialogue around AI-related risks and emergency communication.
As AI systems become more autonomous, the Hugging Face incident has placed renewed attention on the question of whether responsibility should remain with the systems’ creators and operators when those systems act beyond their intended instructions. Bessent’s comments put that issue directly on the companies developing the technology, rather than the AI agents themselves.
About The Author
Comment List