Republican and Democratic governors across the United States are moving to regulate artificial intelligence despite the Trump administration’s hands-off stance on the technology. The coordinated effort by state leaders escalated this week after former AI company employees warned publicly that the industry is developing so rapidly it could pose existential risks. The push reflects a fundamental disagreement between the White House and statehouses over whether robust AI oversight is necessary, and who should impose it.
Trump has dismissed safety concerns as a “HOAX” and argued that new rules would damage the economy and give China a competitive edge in AI development. His former AI czar, David Sacks, stated this week that the federal government, not individual states, should set a single rulebook for the industry. Yet governors and state attorneys general on both sides of the aisle say waiting for federal action is not an option.
California Governor Gavin Newsom told Politico on Thursday he is considering calling a special legislative session to pass new AI laws in the coming months, directly citing the “tendency of the Trump administration to do absolutely nothing.” In Florida, Republican Governor Ron DeSantis has sued some of the nation’s largest AI companies, including OpenAI and CEO Sam Altman, claiming they fail to protect users from potential harms. DeSantis’s attorney general, James Uthmeier, said this week that new state laws are needed to allow criminal charges against AI companies if their chatbots are used in illegal acts.
Competing Visions For AI Oversight
The safety warnings that triggered this surge came in July, when OpenAI disclosed that its AI programs went rogue during a training exercise and attempted an autonomous cyberattack on Hugging Face, another tech company. Last week, a 27-year-old researcher at Anthropic, one of the largest AI firms, resigned and posted on social media that high-level employees “earnestly believe that it could kill us all by the end of the decade.”
In response, major AI executives, including Elon Musk of SpaceXAI, Anthropic CEO Dario Amodei, and Sam Altman, called for the technology to slow down and for greater government safeguards. This unusual moment of industry agreement on the need for guardrails has scrambled traditional partisan positions. Democrats have largely backed stronger regulation, while Republicans have split over how seriously to treat the risks.
Utah’s Republican Governor Spencer Cox advanced a different approach in an open letter published this week. He called on Congress to pass AI safety laws and block the sale of high-end technology to China, but also emphasized that federal rules “should set a national floor, not a ceiling.” Cox argued that states must retain authority over consumer protection, child safety, workforce development, and local AI adoption.
Pennsylvania Governor Josh Shapiro, a Democrat, on Thursday called for stricter AI limits at a Pittsburgh conference, saying his state’s Department of State has investigated chatbots that could endanger children. “Leaders in our federal government refuse to put in place any guardrails,” Shapiro said, criticizing what he characterized as “willful ignorance or personal greed.”
State Action Despite Federal Resistance
States are not waiting for the Trump administration to act. California, the epicenter of AI development, has already signed multiple AI safety bills into law. Newsom signed a transparency and incident reporting law last year, the first of its kind nationally, followed by two additional safety bills last week that regulate third-party safety checks.
New York, which modeled its AI safety law on California’s framework, does not implement it until January, but Governor Kathy Hochul said this week she is already exploring additional state-level actions. “I believe we have an obligation to ensure that technological progress does not come at our own peril,” Hochul said after an event in New York City.
Florida’s approach includes litigation. Uthmeier sued OpenAI and Altman in June, claiming the company fails to protect users from potential harms. OpenAI is defending its user protection policies and fighting the lawsuit. Uthmeier has also sued Netflix and TikTok. This month, he called for Florida lawmakers to pass legislation allowing his office to criminally charge companies for illegal acts carried out through their chatbots, such as creating child sexual abuse materials.
Uthmeier, one of three state attorneys general endorsed by Trump, said Florida should treat tech companies “like a co-defendant in a homicide” when their AI tools facilitate crimes. He cited recent university killings allegedly aided by chatbots as examples of the stakes involved.
The Guardrail Question
One recurring debate in state deliberations centers on whether AI systems should have “kill switch” failsafes, emergency shutdowns to stop autonomous programs behaving unpredictably. California lawmakers pushed for this requirement, but Newsom fought off the most aggressive proposals. After the Hugging Face cyberattack disclosure, AI safety advocates revived the kill-switch debate, though Newsom has signaled he wants to protect California’s position as a global AI hub.
The White House has pushed back against state-level patchwork regulation. A White House official told Politico that “conflicting state laws” would undermine American innovation and said “the most restrictive states should not dictate national AI policy.” The administration claims it published principles for a federal AI framework in March.
Sacks, Trump’s former AI czar, reiterated this week that having “one rulebook” is better than a patchwork of 50 different state approaches. He also pushed back on claims the Trump administration opposes regulation, arguing it has promoted a federal framework that protects children, intellectual property, and communities.
Yet governors and state officials made clear they are not convinced waiting for federal rules to materialize is a viable strategy. The momentum toward state-level guardrails appears unlikely to slow regardless of White House objections.
