Sen. Bernie Sanders is demanding that OpenAI, Anthropic and Meta voluntarily pause development of increasingly powerful artificial-intelligence systems, warning their chief executives that the U.S. Senate could intervene if the companies continue advancing the technology despite concerns about whether humans can reliably control it.
The Vermont independent addressed his warning directly to OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei and Meta CEO Mark Zuckerberg in a letter first shared with Axios. Sanders cited reports of unexpected model behavior, potential biological applications of AI and previous safety commitments made by the companies as grounds for slowing development.
"Mr. Altman, Mr. Amodei and Mr. Zuckerberg: In the interest of humanity, stand by your words. Pause AI development," Sanders wrote. "It is not too late to avoid disaster. Stop building machines that humans cannot control."
Sanders then raised the prospect of government intervention if the companies don't voluntarily change course.
"Let me be very clear: If you do not take appropriate action now, my colleagues and I in the U.S. Senate will," he wrote.
The demand represents an escalation in Washington's debate over frontier AI, where lawmakers are trying to balance concerns about rapidly improving capabilities against pressure to maintain America's technological advantage over China. Broad legislation requiring leading AI developers to halt their work would face substantial political obstacles in the current Congress.
Sanders based part of his argument on previous safety commitments from OpenAI, Anthropic and Meta. According to Axios, the companies have indicated that development could be paused or stopped under circumstances in which increasingly capable systems presented dangers that couldn't be adequately controlled.
The senator argues that recent developments should trigger those commitments.
"At a moment when we have seen human loss of control and the creation of potentially dangerous viruses, your companies are still racing ahead - investing tens of billions of dollars into a technology that nobody can fully understand, predict or control," Sanders wrote.
OpenAI recently slowed the release of its latest model, Astra, while OpenAI, Anthropic and Meta have reported instances of AI systems displaying unexpected or potentially dangerous behavior, according to Axios. Such incidents have become a central focus of AI-safety researchers studying whether more capable models could develop behaviors their creators didn't anticipate.
Biological research has added another dimension to the dispute. Researchers have demonstrated the ability to use artificial intelligence to design functioning viruses that attack bacteria, work that could have legitimate scientific and medical applications.
The research has nevertheless raised concerns about dual-use capabilities. As AI systems become better at biological design, policymakers and researchers are debating whether safeguards can keep potentially dangerous capabilities from being repurposed to assist in developing biological threats.
Pressure for stronger oversight isn't limited to Sanders. More than 1,200 employees at major AI companies have backed calls for the U.S. government to support an international initiative designed to control the pace at which increasingly powerful systems are developed, according to Bloomberg Law.
The employee campaign reflects an emerging divide inside the technology industry. AI companies are spending heavily on chips, data centers, energy and research teams as they compete to build more capable models, while some of the people working inside those companies are asking governments to establish mechanisms for slowing the race if safety research fails to keep pace.
For OpenAI, Anthropic and Meta, voluntarily pausing development would carry significant competitive consequences. Any unilateral slowdown could potentially hand an advantage to domestic competitors or foreign developers that continue training increasingly sophisticated systems.
