The Federal Government Is Becoming an Agent of AI Chaos

House Speaker Mike Johnson at the Capitol on Nov. 3. His chamber is expected to vote on the NDAA next week. (Aaron Schwartz/Bloomberg)

About the author: Michael Kleinman is head of U.S. policy at the Future of Life Institute.

At the behest of a handful of Silicon Valley executives, officials within the Trump administration and their allies in Congress are trying yet again to prohibit states from regulating AI companies. Even as House members heard about the horrific impact of AI chatbots on children, its GOP leadership tried to insert federal pre-emption of state AI laws into the unrelated National Defense Authorization Act. Congress must pass this bill, which is meant to authorize the work of the Pentagon, by the end of the year. Meanwhile, the White House is floating a draft executive order designed to undermine states’ ability to legislate on AI.

Even if the effort to include pre-emption in the NDAA ultimately fails, Republican leadership has been clear that they will try to find other ways to block state AI legislation.

Federal pre-emption was a bad idea when the Senate first proposed and then rejected it 99-1 in July, and it is a bad idea now. Preventing states from legislating on AI wouldn’t only be dangerous for American families and affront to states’ ability to protect their citizens, but also fundamentally bad for business.

Businesses need a stable and predictable regulatory environment. Constant changes in regulation makes it more difficult to plan and impose ongoing compliance costs, the burden of which falls most heavily on smaller businesses that lack large compliance teams. Unfortunately, legislating through defense bills and governing by executive order creates significant instability.

Executive orders can be reversed overnight, as we saw in January when the Trump administration revoked President Joe Biden’s 2023 executive order on artificial intelligence. And the NDAA process bypasses typical committee hearings and expert testimony. Most members of Congress don’t provide input into the NDAA, but must vote yes or no on the final bill handed down by their leadership.

The American public broadly rejects legislating in this manner and resoundingly opposes pre-emption specifically: 57% of all voters, including a plurality of Trump voters, don’t support the proposal. Just 19% do support it. There is no democratic mandate for this approach.

When regulation emerges from hasty backroom deals rather than from robust legislative debate, the result is bad policy that quickly breaks down. One Congress jams contradictory provisions into unrelated legislation, and the next one is left to clean up the mess. This creates the type of uncertainty that suppresses the very innovation that pre-emption advocates claim to protect. Instead, the federal government itself becomes the main source of regulatory chaos.

Meanwhile, states have already begun filling the federal vacuum with thoughtful legislation.

Silicon Valley and its allies like to emphasize that a “patchwork” of more than a thousand state AI laws are introduced each year. But few of these bills ever make it into law, and the vast majority of the select bills that do address real-world harms from AI systems that impact all of us, regardless of political affiliation.

This is precisely why safeguards are emerging not only in blue states like California, but also in red states like Texas and Utah. These laws demand reporting on safety incidents, strengthen whistle-blower protections, and require disclosures on mental health chatbots—all reasonable and targeted interventions that protect our communities and our families.

The current proposals wouldn’t replace these existing state laws with a federal standard. No such standard currently exists, and neither Congress nor President Donald Trump have proposed one. The result would be no guardrails, and crucially, no realistic path to creating them, given Congress’s penchant for gridlock. Indeed, this is likely Big Tech’s deliberate strategy to avoid accountability.

All American companies—from behemoth pharmaceutical companies to the taco truck on your street corner—must comply with federal and state regulations on consumer protection, privacy, and labor standards, just to name a few. That hasn’t stifled American innovation. It has protected the public and boosted consumer confidence in commercial products.

The U.S. remains the most innovative economy on Earth, precisely because regulation creates trust and stability. AI companies deserve no special exemption from new regulations. Federal pre-emption without meaningful federal standards isn’t pro-business policy. It is a gift to a handful of corporations at the expense of market stability, public safety, and democratic accountability.

Congress should reject this approach and do the hard work of crafting actual legislation through regular order. American families and businesses both deserve better.

Guest commentaries like this one are written by authors outside the Barron’s newsroom. They reflect the perspective and opinions of the authors. Submit feedback and commentary pitches to ideas@barrons.com.