GOP Moves Ban on State AI Laws

Another bold move from Congress while you were sleeping last night: an important House committee voted to support a 10-year ban on states enforcing any laws or regulations they adopt for artificial intelligence.

The vote happened at the House Energy & Commerce Committee, which was marking up various pieces of legislation destined to be part of Republicans’ massive tax-and-spending bill. The committee voted along party lines to support the moratorium — all 29 Republicans in favor of it; all 24 Democrats against.

The proposed moratorium, part of a larger “AI and Information Technology Modernization” bill, is straightforward:

No state or political subdivision thereof 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. 

Republicans’ theory here is that a 10-year ban on state regulation of AI will give the federal government more time to develop national standards, presumably in step with the Trump Administration’s policy memo released in April meant to accelerate the development of AI technologies. Of course, in practice a ban would really just buy more time for Elon Musk, Mark Zuckerberg, Sam Altman, and other AI-generated zillionaires to dictate federal AI law; but that’s the rationale Republicans are sticking with.

Obviously a ban on state regulation of AI would have significant implications for corporate compliance, privacy, and cybersecurity teams, as you try to build effective regulatory change and risk management systems — but don’t hold your breath; the proposed ban still has a long road ahead.

First, this particular legislation from the Energy & Commerce Committee needs to be included in the larger tax-and-spending bill for the next fiscal year starting Oct. 1; that will require support from Republican leadership. That tax-and-spend bill then needs to pass both the House and the Senate. Republicans aren’t certain they’ll be able to pass that final bill themselves, which means they might need Democratic help, which means there could still be lots of horse-trading yet to come. That’s typically when 11th-hour surprises happen. 

Second, several political constituencies out there do support states’ rights to regulate artificial intelligence. For starters, Democrats frame the right to regulate AI as pro-privacy and pro-consumer, and they can make plenty of political hay from that stance. During the hearing last night, Rep. Doris Matsu, D-Calif., labeled the moratorium “slap in the face to American consumers.”

U.S. states themselves will also want to guard their freedom to enact laws and rules as they see fit, and that cuts across partisan divides. For example, the two states most known for already enacting laws to regulate artificial intelligence are Colorado and California — but Utah just did the same, with new requirements for when businesses must tell consumers that they’re interacting with an AI chatbot. Nobody would put Utah in the same progressive vanguard as Colorado and California.

All that said, clearing the committee markup is an important first hurdle for the moratorium. Moreover, it has a crucial backer in the Senate: Ted Cruz, chairman of the Senate Committee on Commerce, Science, and Transportation. That counts for a lot, because the Senate is where lots of legislation (both good and bad) gets killed.

Would an AI Ban Even Matter, Really?

That’s a good question. For the largest of businesses, you’re already going to be subject to the constraints of the EU AI Act. So even if Republicans in Washington stop U.S. states from regulating privacy, security, and algorithmic discrimination issues — well, you’re going to have to confront all that anyway thanks to the EU law. 

Sure, a ban against U.S. states would ease the burden around the edges of your compliance burden; but you’ll still need the same fundamental capabilities — say, an ability to audit the results of an AI process to detect AI-driven discrimination, or a catalog of use-cases for AI at your business — that you were always going to need.

States will also try to sidestep this ban by extending existing consumer protection and anti-discrimination laws to include AI, rather than adopting new legislation specific to AI. 

For example, California state attorney general Robert Bonta issued an advisory in January warning companies about possible abuses of AI. In that advisory he essentially said that California state law forbids discrimination, regardless of whether the discrimination comes from humans, AI, or any other system. Other state attorneys general have made similar statements. 

Would this proposed moratorium bar states from enforcing those pre-existing laws, too? It seems not; the text says that the legislation can’t be construed to bar enforcement of other “generally applicable laws” that impose duties of testing, documentation, or civil liability on other systems that don’t use AI. 

So I wouldn’t be surprised to see clever state legislators or attorneys general tweaking state laws in just the right way to keep AI systems under their watchful eye. As always, compliance burdens march on.