Tucked away in President Donald Trump’s “big, beautiful bill,” which recently passed the House of Representatives, is an unforeseen Trojan horse. It is not even clear if Trump is aware of this last-minute insertion by House Republicans, but one can only hope he is made aware and goes on record as opposing it. It goes completely against his efforts to help make the internet a safer place for everyone, especially our children.
Buried in Section 43201(c) of the 1,118 page budget bill is one little sentence: “Subsection (c) states that no state or political subdivision may enforce any law or regulation regulating artificial intelligence models, artificial intelligence systems, or automated decision systems during the 10-year period beginning on the date of the enactment of this Act.”
What incredible mischief that one little sentence produces.
For example, Utah has been at the forefront of policy efforts to shield children from the worst effects of our internet-centered culture. It was one of the very first states to insist that online porn sites implement age verification measures, and this year Utah required that app stores must also verify users’ age.
Utah also legislated protections for its citizens when they interact with so-called “mental health chatbots.” Utah’s stance can only be applauded — but the clause in the federal budget bill would imperil the implementation of all of the legal protections put in place by the state. Utah would be prohibited by the federal government from enforcing any of its AI-related laws if this sentence were to be preserved in the reconciliation version of the bill.
As co-editor and contributor to "The Oxford Handbook on AI Governance," I condemn this mischief in the strongest possible terms. This is not simply irresponsible, but actually prevents state governments from acting responsibly to safeguard its citizens from the increasingly visible downsides of unfettered AI deployment. In that sense, the insertion of this one sentence is patently malicious.
It is clear that our tech overlords would like nothing better than to be unfettered in pursuit of their objectives for AI, regardless of any harm that might come to Americans from that pursuit. As one commentary by J.B. Branch and Ilana Beller put it, “The provision is a message to the states: Sit down and shut up while Big Tech writes its own rules or continues to lobby Congress to have none at all.”
And that failure of Congress to write any rules at all, except the Take It Down bill championed by Trump and his wife, Melania, is equally unconscionable. States have been forced into action by Congress’ stubborn inaction. And states have rightly stepped into that vacuum.
In addition to Utah, there are numerous states that have tried to hold AI-promoting companies responsible for the behavior of their products. Various states have laws against AI-generated deception that might influence elections; other states require citizens be notified when a decision about, say, mortgage approval, has been made by AI and that they be provided recourse to contest such a decision. Some states prohibit AI voice and image cloning of citizens; other states prohibit medical insurance denials by AI; numerous states have legislation providing standards for driving by autonomous vehicles, and so on and so forth. Which of these would anyone wish to put under federal moratorium?
The rejoinder, of course, is that AI companies feel hamstrung by the patchwork of state laws that currently exists. Tough, I say. Congress must finally step up to the plate, rather than forfeiting the game. Maybe tech companies’ bellyaching about that patchwork is all that will spur Congress to get the job done. Given Congress is incapable of delivering more than a lowest common denominator on AI, even if Congress acts, states will have to be the ones to raise the bar that Congress won’t.
And that’s good. State-level democracy has shown itself to be much more vibrant than what currently exists at the federal level. Let the states be the laboratory where the best practices in AI regulation can be honed and then copied nationwide. AI absolutely needs guardrails; if Congress is currently impotent, let the states do the job that the federal government seemingly cannot. The 10th Amendment indicates states are well within their right to do so.
And make no mistake: those guardrails are needed more than ever as AI capabilities increase. There are new AI-based threats coming into focus for which we dare not wait another decade to tackle. The family of Sewell Setzer III, a 14-year-old boy who shot himself after an AI chatbot encouraged him to do so, has just won a major court victory against Character.AI. The company protested it should be protected from liability on a First Amendment claim, a claim a federal judge has now dismissed, enabling the family’s wrongful death suit against the company to move forward.
One of the directors of the AI Now Institute summed up the situation well: “The proposal for a sweeping moratorium on state AI-related legislation really flies in the face of common sense. We can’t be treating the industry’s worst players with kid gloves while leaving everyday people, workers and children exposed to egregious forms of harm.”
Amen to that. All senators and congresspersons, regardless of their partisan affiliation, should work together to strip this one sentence from the reconciliation budget bill. Contact yours right away.