TOPSHOT – Journalists run through the U.S. Supreme Court Plaza while delivering their opinions to the news media… [+] A correspondent photographs the court as it issues its ruling in Washington DC on July 1, 2024. (Photo: Drew Angerer/AFP) (Photo: Drew Angerer/AFP via Getty Images)
AFP via Getty Images
A significant blow to regulatory powers dealt by the Supreme Court earlier this month could have major implications for nascent efforts to regulate artificial intelligence in the United States.
In Roper Bright Enterprises v. Raimondo, the Supreme Court overturned the landmark 1984 case Chevron v. Natural Resources Defense Council. The Chevron doctrine required that when Congress had not directly addressed the central issue in a dispute, courts were required to defer to the relevant agency’s interpretation.
So when it comes to AI regulation, the Roper-Bright decision is a clear victory for those who believe overregulation is the greatest danger to AI development, but it’s far from decisive — and a loss for those who believe the U.S. should pursue AI governance rather than thwart it.
In the wake of the Roper-Bright decision, it will be up to Congress to clearly define the scope of authority that will be granted to any future designated AI agency (or to enact legislation granting existing agencies significant powers to regulate AI).
“The loss of discretion in Chevron makes it clear that it’s the courts, not the regulators, who decide the scope of their authority,” Jeremy Richardson, a partner at the law firm Smith, Gambrell & Russell, said in an interview, adding that the decision doesn’t favor either party because federal judges represent conservative and liberal views.
Additionally, Richardson noted that the Administrative Procedure Act, which governs agency rulemaking and adjudication procedures, remains in effect. Loper said he is not overriding any enabling statutes. Instead, he makes it clear that if there is a question about the scope of authority granted by an enabling statute, that is a legal question for the courts, not the regulatory agencies.
As a result, Richardson believes a post-Chevron world could see the emergence of an empowered AI regulator: “If an agency were chartered to oversee AI, if that happens, it would have broad (and, in light of Roper’s opinion, clear) powers, and I would hope that smart, well-informed people would respond to the call to establish appropriate guardrails,” he said.
But Justice Elena Kagan’s dissent took a different view on the extent to which Roper Bright limits the consideration of expertise.
“Congress knows it cannot, and indeed cannot, write flawless regulatory law,” she wrote. “Congress knows that such law will necessarily contain ambiguities that other parties must resolve and gaps that other parties must fill. And Congress wants those parties to be accountable agencies, not the courts.”
The impact of the disappearance of Chevron deference will therefore depend on whether courts can fairly evaluate issues that are beyond their skills and training. For most justices, regulatory issues related to science and technology (the most pressing of which is artificial intelligence) fit precisely into this category. Justice Kagan’s dissent argues that certain issues that may arise in regulatory contexts involve technical subject matter, so those responsible for interpreting them must have sufficient relevant knowledge. “Government agencies have expertise in those areas; courts do not,” she wrote.
WASHINGTON, DC – February 4: U.S. Supreme Court Chief Justice John Roberts and Chief Justice… [+] Justice Elena Kagan attends the State of the Union Address on the floor of the U.S. House of Representatives in Washington, DC, February 4, 2020. (Photo by Mario Tama/Getty Images)
Getty Images
Few would dispute Kagan’s explanation of the benefits of deferring to Chevron, said Charlie Block, a research fellow at the Institute for Law and AI. But “different people have different views on the costs,” he said in an interview. Where you look depends a lot on whether you’re worried about under- or over-regulating AI.
Block said AI regulation is a “classic example” of Chevron’s deference for two main reasons: 1) Regulators targeting AI have the benefit of deep expertise, and 2) regulatory flexibility is paramount given the rapidly changing landscape around AI: government agencies are more likely to update rules more frequently and quickly than Congress.
While the Loper-Bright decision does not completely strip agencies of their power, it does raise the bar for detail regarding future legislative language defining agency responsibilities. This change may put pressure on Congress to spell out with absolute specificity which agencies are responsible for what types of AI regulation.
In other words, the ruling stops short of barring Congress from delegating important issues to regulators. Mark McCarthy, an adjunct senior fellow at the Brookings Institution’s Center for Technology Innovation, says the ruling could have been much more extreme. “The way the Court overturned Chevron leaves agencies with ample policy space to craft measures that protect the public interest going forward, provided Congress clearly crafts the delegation of authority,” he wrote.
This is a pretty big caveat: Lawmakers have not always clearly delegated regulatory authority, and the increasing political polarization of the Legislature is unlikely to make that any more likely.
Additionally, Chevron’s application could change depending on a variety of factors, including who is in the White House. For example, if Trump is re-elected and follows through on his plan to make most federal jobs politically appointed (as he began doing in 2020 through “Schedule F” — a recommendation in the Heritage Foundation’s Project 2025 plan for Trump’s second term), Roper could actually help shape AI regulations, Block explained.
Current political, legal, and technological realities suggest that Roper Bright will strike a blow for AI regulation, but the political utility of deferring to Chevron will always be context-dependent.
“If Chevron has the backing of government agencies but not the backing of the courts, everybody likes Chevron. And if Chevron has the backing of the courts but not the backing of government agencies, everybody hates Chevron,” Block said.
For now, big U.S. AI companies seem to be getting a pretty positive reception from all quarters. But Block said, “There remains a lot of uncertainty, especially in lower courts, about what exactly Roper Bright means and how significant it is.” Government agencies may hesitate to implement ambitious regulations if they’re not sure they’ll be at risk of litigation, Block said.
“We’ve been living in a Chevron world for 40 years,” Block said. It’s an open question how much power government agencies would enjoy under the weaker Skidmore Act deference doctrine. So a Supreme Court decision to overrule Chevron would unravel what Block called “an entire universe of previously settled case law.”
He expects regulators to tread cautiously, and Silicon Valley AI companies have signaled they don’t intend to follow suit.
