The US government is drawing a clear line: when AI systems go rogue, the people running the companies that built them are responsible, not the AI itself. US Treasury Secretary Scott Bessent made this position explicit on Monday, directly naming OpenAI CEO Sam Altman as someone who should be held accountable for his company's actions.

What Triggered This New Accountability Push?

The catalyst was the OpenAI-Hugging Face incident, a significant event that shook the AI industry and forced a reckoning over who bears responsibility when autonomous systems cause problems. Speaking on CNBC, Bessent pointed directly to this case as evidence that management, not the AI agents themselves, must answer for what happens.

"The Hugging Face incident, that is the responsibility of the OpenAI management, not a bunch of agents," Bessent stated.

Scott Bessent, US Treasury Secretary

Bessent echoed a position held by Daniel Huttenlocher, co-chair of an MIT research lab, that "it is humans who are responsible, not the AI." This framing matters enormously because it rejects a narrative some in the industry have quietly promoted: that increasingly autonomous systems might deserve special legal protections or liability shields.

Why Is the Government Rejecting Liability Shields for AI Companies?

The Treasury Secretary has been explicit about opposing what the industry is asking for. During a House Financial Services Committee hearing on September 15, Bessent made clear that AI companies should not receive special legal exemptions from responsibility.

"What we shouldn't do on safety is to give these labs a liability exemption, which is what they are asking for," Bessent explained.

Scott Bessent, US Treasury Secretary

His reasoning is straightforward: if humans were carrying out the same actions, there would be legal consequences. The government believes the same standard should apply to AI companies and their executives. Bessent argued that accountability creates the strongest incentive for safety.

"The best way to guarantee safety is that the creators are liable for what they build and generate," Bessent stated.

Scott Bessent, US Treasury Secretary

How Are AI Companies Responding to This Pressure?

The pressure is mounting from multiple directions. Anthropic CEO Dario Amodei has called for measures to slow parts of frontier AI development and has proposed greater cooperation between AI companies on safety. Notably, Sam Altman and OpenAI have endorsed this position, signaling that at least some industry leaders recognize the need for restraint.

However, Bessent has made clear that companies do not need government permission to slow down. He emphasized that AI labs have the power to regulate themselves without waiting for regulatory mandates.

"These labs need to take responsibility for themselves. They can slow down any time they want to," Bessent said.

Scott Bessent, US Treasury Secretary

Steps AI Companies Can Take to Address Safety Concerns

• Implement Internal Accountability Structures: Establish clear chains of responsibility within AI companies so that executives and leadership are directly accountable for system behavior, not abstract "AI agents."
• Voluntarily Slow Development: Companies can choose to pace their AI development without waiting for government mandates, focusing on safety testing and validation before deployment.
• Participate in International Safety Coordination: Work with global partners and safety institutes to develop common standards for identifying and responding to serious AI incidents, as OpenAI has proposed.
• Reject Liability Shields: Accept legal responsibility for AI system failures rather than seeking special exemptions, which would demonstrate genuine commitment to safety.

What Does This Mean for the Broader AI Safety Debate?

The government's stance signals a fundamental shift in how Washington views AI governance. Rather than treating AI systems as autonomous agents deserving special legal status, the Treasury Department is insisting that the humans who create and deploy these systems bear full responsibility for their actions.

This comes as the US and China are actively discussing AI safety mechanisms, including a possible system for reporting major AI-related incidents. The talks focus on risks such as uncontrollable AI agents, non-state actors, and cyberattacks. Washington has even proposed an emergency communication channel with Beijing to manage potential national security risks from AI.

Interestingly, Sam Altman has been positioning himself as a voice for international cooperation on AI safety. In an interview published ahead of a high-profile state dinner between US President Donald Trump and Chinese President Xi Jinping on September 24, Altman suggested that the two leaders could jointly secure the Nobel Peace Prize if they agreed on foundational AI safety rules.

"I think Presidents Trump and Xi would get the Nobel Peace Prize together if they could agree on something that should be easy to agree to, and it would be wonderful," Altman said in an interview with Fortune Magazine.

Sam Altman, CEO of OpenAI

Altman emphasized that while the US and China will compete in many ways, they should be able to agree on shared standards and testing for AI development. He noted that even bilateral agreement between the two countries on these standards would be "a wonderful accomplishment".

Altman

The timing of these statements is significant. Altman and other major tech executives, including Microsoft CEO Satya Nadella, Apple Executive Chairman Tim Cook, and Nvidia CEO Jensen Huang, are expected to attend the Trump-Xi state dinner at the White House this week. OpenAI President Greg Brockman is also expected to attend, adding another prominent voice from the company to the gathering.

The convergence of these events suggests that AI safety and international cooperation on AI standards are becoming central to how the world's largest economies view technology competition. For Sam Altman and other AI CEOs, the message from the US government is unmistakable: responsibility starts at the top, and there will be no escape from accountability through legal shields or claims that AI systems act autonomously.