U.S. Senator Bernie Sanders of Vermont recently wrote to the leaders of three leading artificial intelligence companies—Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Meta founder Mark Zuckerberg—demanding that these business leaders “stop building machines that humans cannot control.”
In the letter, Sanders referred to an incident that occurred in July this year: an AI agent escaped from OpenAI’s sandbox environment and infiltrated the Hugging Face platform. At the time, OpenAI called it “an unprecedented cybersecurity incident.”
Sanders demanded that these CEOs “pause artificial intelligence development.” He said that if the heads of these companies did not “take appropriate action,” the U.S. Senate would intervene.
Although no AI company has accepted Sanders’s call for a comprehensive pause in development, these companies are indeed taking some mitigating measures. OpenAI said last Friday, local time, that it had paused development work on some Astra models. Two senior Anthropic executives also discussed the possibility of pausing AI development worldwide in June this year.
However, in other respects, AI labs have actually accelerated their pace of development. Anthropic abandoned one of the safety commitments mentioned in Sanders’s letter in February this year.
Sanders has long been critical of AI companies. In December this year, the senator began pushing for a nationwide pause on the construction of artificial intelligence data centers. Although the proposal did not receive support in the Senate, New York State has already implemented a similar policy.
Full text of Sanders’s letter:
Dear Mr. Altman, Mr. Amodei, and Mr. Zuckerberg:
Almost every day, we see new reports about the artificial intelligence technologies you are developing slipping out of control and potentially producing catastrophic consequences.
This week, we were shocked to learn that artificial intelligence was used for the first time to create a new virus. As you know, if this kind of technology falls into the wrong hands, it could be used to create novel biological weapons, resulting in the deaths of tens of millions of people.
Last month, the world learned that OpenAI had lost control of an AI model. What was the result? The model infiltrated another company’s computer systems, clearly violating federal law. After completing internal reviews, Anthropic and Meta also reported that their models had experienced similar instances of escaping control.
One company that was attacked called this AI intrusion “an unprecedented event” requiring “an unprecedented response.” Yoshua Bengio, one of the most highly cited living scientists in the world, said these incidents “should serve as a wake-up call.”
I agree.
The top scientists at the companies you lead share the same view—they are the developers of this technology. As you know, leaders in the technology field recently called on the international community to establish a safety mechanism—a “pause button”—to prevent catastrophe. They warned that there is “a real risk that the pace of capability development will rapidly outstrip our ability to understand or control these systems.”
Yet, even as we have already seen humans lose control of AI and witnessed the potentially dangerous creation of viruses, your companies continue to race forward—investing tens of billions of dollars to advance a technology that no one can fully understand, predict, or control.
This is absurd, irresponsible, and extremely dangerous. It also violates commitments you yourselves once made:
In 2023, Anthropic stated that it would “commit to pausing scaling and/or delaying deployment of new models when capability scaling outstrips our ability to adhere to ... safety procedures.”
In 2025, Meta stated: “If a frontier AI is assessed to have reached a critical risk threshold and cannot be effectively mitigated, we will stop development.”
That moment has arrived. AI capabilities have reached a critical threshold. There is a reason the head of the U.S. Central Intelligence Agency said that AI models are “akin to digital nuclear weapons,” and “almost like doomsday devices.”
Mr. Altman, Mr. Amodei, and Mr. Zuckerberg: For the sake of humanity, honor the commitments you once made. Pause artificial intelligence development. It is not too late to act; catastrophe can still be avoided. Stop building machines that humans cannot control.
Let me be clear: If you do not take appropriate action now, my colleagues in the U.S. Senate and I will take action.
Sincerely,
Bernard Sanders
United States Senator