The 2024 Election Results & AI Policy: Here Comes the Mother of All State Regulatory Patchworks

For artificial intelligence (AI) policy in the United States, the 2024 election results likely mean state-level regulatory legislation will accelerate rapidly as key states adopt the same “we-gotta-do-it-to-save-humanity-since-Congress-won’t” rhetorical ploy we already saw at work in California this summer with the bill that Gov. Newsom vetoed (SB 1047).

But the state AI bills that will advance now won’t look like that California one at all. The new wave of state AI bills will instead resemble the “algorithmic discrimination” bill that passed in Colorado in May, almost passed in Connecticut as well, and is being floated in Texas currently.

The high-level branding of these new state measures will shift from “AI is unsafe” to “AI is biased,” and the regulatory advocates currently working hard at the federal level will mostly migrate to the halls of a half dozen key statehouses. They will begin lobbying hard for the model legislation being developed through the new Multistate AI Policymaker Working Group, which is now being coordinated by the Future of Privacy Forum. They basically want to replicate the success they had in Colorado.

However, now that so much of the action will be at the state level, other special interests will take notice of these measures and seek to hitch a ride. These new state AI measures will grow to be more bloated bills, with additional provisions potentially being added to satisfy copyright, child safety, and union / anti-automation interests, just to name a few.

Rise of the AI Legion of Doom

Thus, a veritable “AI Legion of Doom” will now align to push these mega-AI state bills. The old “AI Safety / X-risk” crowd behind the California bill may even seek to get in on the action by rebranding as being part of the “AI Ethics” movement (although it won’t likely work). Ironically, all that added interest and resulting legislative bloat may be the only thing that slows or stops these mega-measures in some states. The problem with creating legislative Christmas trees and placing presents underneath for every special interest is that eventually the tree often falls over under its own weight. Nonetheless, such mega-measures could get through in a few particularly blue states.

Of course, many other narrower measures will advance, too, as the mega-measures soften the ground for other legislative strikes. Recall that, as we approach the end of the year, an astonishing 750 AI bills are pending currently across the nation, and dozens have already passed. As I have noted many times before, this is an unprecedented volume of regulatory activity. No other emerging technology in U.S. history has ever faced this sort of legislative tsunami before the technology has even enjoyed widespread diffusion.

In any event, unless a serious counter-movement develops to stop it, I expect 5 or 6 states to pass comprehensive AI regulatory measures by this time next year, and dozens (perhaps hundreds?) of narrower AI bills will get enacted alongside them as this mother of all state regulatory patchworks spreads. The compliance costs will be extraordinary and absolutely crush AI innovation and competition in the United States. It will be an unmitigated disaster.

Source: Multistate

Why Federal AI Preemption Likely Won’t Happen

Some people might expect the new Congress and returning Trump administration to preempt all this state level activity in some fashion, but that’s going to be very hard for two reasons.

First, it is quite complicated to comprehensively preempt all this state AI regulatory meddling when the bills claim to simply be regulating “bias” and consumer harm. States will claim that is their prerogative even though their regulatory actions will have extraterritorial effect. Digital bits and algorithms don’t stop at state borders, after all. But regulatory advocates won’t worry too much about the free flow of digital speech and commerce; they will instead seek to divide-and-conquer the AI ecosystem in the same way they have long wanted to balkanize the Internet. Congress stopped that nightmare scenario in the mid-1990s by preempting state and local regulatory meddling with digital commerce and online speech. But that was actually much easier to do than widespread AI preemption would be because AI is perhaps the most general of all general-purpose technologies. AI’s boundaries are extraordinarily amorphous — both technically and legally. And, so, state and local governments will find far more ways to regulate all things algorithmic, and Congress will struggle to devise a sensible way to stop them.

The second reason that AI preemption is unlikely to happen comes down to politics: Many Republicans will resist the idea of limiting what states can do on this front based on “States rights” thinking that remains too firmly planted in the agrarian era. As noted, digital speech and commerce isn’t like a field or factory that sits within neatly defined geographic confines. Digital activity transcends lines on a map. It’s everywhere. The same is true of algorithmic speech and commerce. But a huge number of Republicans today still think in 19th Century terms when it comes to what constitutes “interstate commerce” and have a sentimental attachment to antiquated notions of “States rights” that simply do not match up with modern technical or economic realities. The fact is, state and local governments have absolutely no business regulating 99% of the things that they seek to control through most of these new AI bills. Moreover, had state and local government applied similar thinking to Internet speech and commerce, America would have never witnessed the explosion of online activity and speech we enjoyed with the rise of the Digital Revolution.

But few people today remember the policy origins of the Digital Revolution and the crucial role that preemption played in ensuring that America became the unparalleled global leader on this front. Instead, many policymakers are prepared to throw it all away and let AI markets and algorithmic speech now be encumbered with a “death-by-a-thousand-cuts” state-by-state (and even city-by-city) regulatory regime for artificial intelligence and advanced computation. I’d love to see Congress again make a stand for technological freedom and digital free speech, but for the reasons I’ve identified here, I think it’s very unlikely.

So, 2025 will be the year that the AI Legion of Doom makes its stand in the states. Are you prepared for the many policy battles ahead? The freedom to code, freedom to compute, freedom to innovate, and freedom to speak are at risk. I look forward to seeing you in statehouses across the nation as this fight gets underway.

Additional Reading:

• Adam Thierer, “Texas should reject the European path toward AI regulation,” R Street Analysis, Oct. 31, 2024.
• Adam Thierer, “Colorado Opens Door to an AI Patchwork as Congress Procrastinates,” R Street Analysis, May 20, 2024.
• Adam Thierer, “California Rejects AI Regulatory Extremism,” R Street Analysis, Sept. 30, 2024.
• Adam Thierer, “State and local meddling threatens to undermine the AI revolution,” The Hill, January 21, 2024.
• Neil Chilson & Adam Thierer, “A Sensible Approach to State AI Policy,” Federalist Society Regulatory Transparency Project blog, Oct. 9, 2024.
• EVENT: “What Should the Next Administration Do About AI?” Federalist Society Regulatory Transparency Project, [featuring Ash Kazaryan, Adam Kovacevich, Neil Chilson, and Adam Thierer moderating], Oct. 15, 2024.
• PODCAST: “California’s AI Law SB 1047 & What It Means for Tech — Adam Thierer,” SVIC Podcast, Sept. 10, 2024.
• Adam Thierer, “W​here Does California’s AI Bill (SB 1047) Fit Historically?” Medium, Aug. 26, 2024.
• Adam Thierer, “The Policy Origins of the Digital Revolution & the Continuing Case for the Freedom to Innovate,” R Street Real Solutions, Aug. 15, 2024. [+ video of panel talk]
• Adam Thierer, “California and Other States Threaten to Derail the AI Revolution,” R Street Analysis, May 2, 2024.