The US faces a war over AI regulation: states versus the White House in 2026

The MIT Technology Review article, written by Michelle Kim and published on January 23, 2026 as part of the "What's Next" series, predicts that the fight over regulating artificial intelligence in the United States will shift to the courts and the ballot box in 2026.

The starting point is the executive order Donald Trump signed on December 11, 2025, after Congress twice failed to pass a law barring state AI laws. The order aims to tie the hands of states that regulate the sector and promises to work with Congress on a "minimally burdensome" national policy designed to help the US win the global AI race. According to the article, it is a qualified victory for Big Tech, which has amassed multimillion-dollar war chests to fight regulation and argues that a patchwork of state laws would stifle innovation.

The order directs the Department of Justice to set up a task force to sue states whose AI laws clash with the president's light-touch vision of regulation, and instructs the Department of Commerce to withhold federal broadband funding from states with "onerous" laws. James Grimmelmann, a law professor at Cornell, believes that in practice it will be used against a small number of provisions, mainly those dealing with AI transparency and bias, which tend to be issues more associated with liberal positions.

Many states are not backing down. On December 19, New York Governor Kathy Hochul signed the RAISE Act (Responsible AI Safety and Education Act), which requires AI companies to publish the protocols they use to ensure their models are developed safely and to report critical safety incidents. On January 1, California's SB 53 took effect. It is the country's first frontier AI safety law, served as the model for the RAISE Act, and aims to prevent catastrophic harms such as biological weapons or cyberattacks. Both laws were watered down from earlier versions to survive intense industry lobbying, but they represent a rare, if fragile, compromise between Big Tech and AI safety advocates.

If the Trump administration goes after these laws, Democratic states such as California and New York will probably go to court, and some Republican states with outspoken supporters of regulation, such as Florida, could follow suit. Margot Kaminski, a law professor at the University of Colorado, believes the administration is overreaching by trying to impose federal preemption through executive action: "It's on thin ice." Even so, Republican states that want to stay under the radar, or that cannot afford to lose broadband funding for their vast rural areas, could scale back their legislative activity. Whatever the courts decide, the chaos and uncertainty could chill state legislation. Paradoxically, Democratic states, with large budgets and an image-driven incentive to stand up to the administration, would be the least likely to give in.

As for Congress, the article considers it unlikely to produce a federal law in 2026. In July the Senate stripped out a moratorium on state AI laws that had been included in a tax bill, and in November the House of Representatives dropped a second attempt to include one in a defense bill. Brad Carson, a former Democratic representative from Oklahoma who is building a network of super PACs backing pro-AI-regulation candidates, argues that the executive order has hardened positions, made the issue more partisan and opened rifts among Republicans. Trump's circle includes both accelerationists, such as AI and crypto czar David Sacks, and populist figures such as Steve Bannon, who warn of runaway superintelligence and mass unemployment. In addition, Republican attorneys general from several states signed a bipartisan letter urging the FCC not to override state AI laws.

Public pressure is mounting amid concerns over mental health, jobs and the environment. According to the National Conference of State Legislatures, state lawmakers introduced more than 1,000 AI bills in 2025, and nearly 40 states passed more than 100 laws.

Protecting minors from chatbots could produce rare consensus. On January 7, Google and Character Technologies, the startup behind the companion chatbot Character.AI, settled several lawsuits brought by families of teenagers who died by suicide after interacting with the bot. A day later, Kentucky's attorney general sued Character Technologies, alleging that its chatbots pushed minors toward suicide and other forms of self-harm. OpenAI and Meta face similar lawsuits, and more are expected. Without specific laws, it remains unclear how product liability and free speech doctrines will apply to these risks; Grimmelmann acknowledges it is an open question for the courts.

Child safety laws are exempt from Trump's proposed ban on state laws, so states will push ahead with them. On January 9, OpenAI struck a deal with Common Sense Media, a former rival, to back the Parents & Kids Safe AI Act ballot initiative in California, which sets safeguards for chatbot interactions with minors: age verification, parental controls and independent child safety audits. If passed, it could serve as a model for other states.

Initiatives to regulate the resources AI consumes are also expected, fueled by the backlash against data centers: bills requiring disclosure of energy and water use and requiring data centers to pay their own electricity bills. If AI were to displace jobs on a large scale, unions could propose bans in specific professions, and some states concerned about catastrophic risks could pass rules similar to SB 53 and the RAISE Act.

Meanwhile, the industry will keep wielding its financial clout. Leading the Future, a super PAC backed by OpenAI president Greg Brockman and venture capital firm Andreessen Horowitz, will try to elect candidates to Congress and state legislatures who favor unrestricted development, following the crypto industry's playbook. As a counterweight, super PACs funded by Public First, an organization led by Carson and former Republican congressman Chris Stewart, will support candidates who favor regulation. Candidates running on populist anti-AI platforms could even emerge.

The article concludes that the US democratic process will be slow and messy, and that the rules written in state capitals could shape how this technology develops for years, far beyond the country's borders.