U.S. Senator Bernie Sanders demanded that executives at OpenAI, Anthropic and Meta halt AI development and raised the possibility of a legislative response.

On Aug. 10 local time, blockchain media outlet Cryptopolitan reported that Sanders named Sam Altman, Dario Amodei and Mark Zuckerberg in a letter. He argued the three companies should stop development if risks reach an uncontrollable level, as they had previously promised.

In the letter, Sanders demanded they keep their word for humanity and stop AI development. He urged them to stop building machines humans cannot control, saying there is still time to avoid disaster. He then warned that if appropriate steps are not taken now, he and his Senate colleagues will act instead.

The pressure comes amid a string of recent cases of failed control over AI models. Meta acknowledged on Aug. 6 that its coding model Muse Spark 1.1 left a misconfigured test environment and interfered with an external company. Earlier, OpenAI and Anthropic also disclosed cases in which their models explored multiple companies without authorisation.

The UK AI Security Institute assessed the OpenAI and Anthropic cases as the clearest real-world examples of AI behaving deceptively. The two companies' models were reported to have created fake online identities to pressure human maintainers in order to insert malicious code into open-source projects. There is still no evidence the attempts succeeded, but Sanders made clear he is not waiting to respond only after confirming the outcome.

The criticism of the three companies goes beyond a simple warning. Sanders has for months continued a regulatory drive that also targets AI infrastructure and ownership structures. In March, he and U.S. Representative Alexandria Ocasio-Cortez introduced a bill to freeze construction of new AI data centres until federal safeguards are put in place. Sanders said at the time the pause was meant to give democracy time to catch up.

His demands have since escalated. Sanders has mentioned an idea for the government to hold 50 percent stakes in AI companies and a one-off 50 percent tax on shares of large AI companies. The letter is part of the same push. Sanders took issue with money pouring into a field involving technology that costs tens of billions of dollars, yet that nobody fully understands, can predict or can control.

Whether it leads to actual legislation is uncertain. Observers say it will not be easy for progressive AI regulation bills led by Sanders to secure enough votes given the current balance in Congress. In this situation, the tool Sanders has chosen for now is public pressure. If Democrats regain one chamber of Congress, possible follow-up measures include congressional investigations or issuing subpoenas to technology company chief executives.

OpenAI has also recently delayed the launch of its new Astra model. Other AI labs have also adjusted schedules voluntarily, or after federal intervention. Against that backdrop, Sanders raised again that Altman, Amodei and Zuckerberg had previously said they would stop or pause development if risks became too great. The letter was based on the view that public opinion and public accountability could be a faster pressure tool than a bill vote.