Gene J. Puskar/AP Photo
A self-driving tractor trailer from Aurora Innovation maneuvers around a test track in Pittsburgh, March 14, 2024.
The House spending bill’s proposal to bar states from regulating artificial-intelligence systems for a decade stands to erase the only laws in America governing the use of driverless cars—vehicles that have been responsible for thousands of crashes and multiple deaths over the last four years, federal data shows. “Autonomous vehicles,” or AVs, use artificial-intelligence technology to run; there are currently no federal regulations for their use, and traffic rules more generally are typically governed at the state level. So without state laws, AV use would be unregulated.
“If this provision moves forward, some companies may be more conscientious about safety than others, but we always have to consider the lowest denominator,” said Advocates for Highway and Auto Safety President Catherine Chase. “It’s really quite lucky that there haven’t been more fatalities and injuries considering the lack of rules and regulation.”
House lawmakers passed the provision as part of their mega-spending bill last month. It is now with the Senate, where lawmakers are expected to make substantial changes. While the bill’s marquee provisions relate to gutting Medicare, Medicaid, and other programs to fund tax cuts for the rich, it also includes multiple instances of fine-print maneuvering designed to enrich the wealthy and benefit industry, as the Prospect has chronicled.
More from Whitney Curry Wimbish
The provision banning state AI regulation is among its many giveaways to Silicon Valley, which also include $3 billion worth of line items to integrate artificial intelligence into the military and border control. As part of a separate $500 million appropriation to the Commerce Department to use AI to modernize its information technology, the provision purports to ensure the success of that effort by forbidding “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.” Advocates and legal experts said the provision would also repeal existing laws.
Thirty-four states and the District of Columbia have enacted autonomous vehicle regulation as of the end of last year. The laws vary significantly between jurisdictions and outline a range of requirements. In Tennessee, for example, the manufacturer owner of an autonomous vehicle must maintain primary automobile liability insurance providing at least $5 million for death, bodily injury, and property damage. Among California’s multiple laws for AVs is one requiring that manufacturers maintain a dedicated emergency telephone line.
The Trump administration has been loosening federal regulations on AVs, just as the proposal seeks to roll back state ones. For instance, companies will soon no longer be required to disclose the same level of detail about driverless car crashes as they have for the last four years. Under amendments to the National Highway Traffic Safety Administration’s 2021 standing general order on crash reporting, companies will no longer be required to report when an AV or driver-assisted car crash “involved” a vulnerable road user—just when the car hits that person. They’ll no longer need to always report when such a vehicle damages property—just when the damage is expected to exceed $1,000.
The Trump administration has been loosening federal regulations on AVs, just as the proposal seeks to roll back state ones.
The amendments also eliminate routine monthly reporting requirements and extend the time by which reports must be made. Additionally, it will allow companies to redact more information because they consider it “proprietary,” said one former NHTSA official who spoke on condition of anonymity. All of this was a boon to driverless car manufacturers even before the House spending bill, Chase and others said. The changes take effect June 16.
The data reported to the government over the last four years describes thousands of driverless car crashes, malfunctions, injuries, and multiple deaths. Crashes have increased steadily since 2021. There were 544 driverless car crashes last year, the database shows. In the first four months of this year, there’s already been well over half that.
“We do expect, unfortunately, that autonomous vehicles are going to be more widespread,” Chase said. “The roads will get more dangerous … There has been a push for AVs to be on the roads because there’s a lot of money being invested into research and development, but they by and large have not been monetized on a wide-scale basis, and there’s pressure to start making money.”
The world’s wealthiest person, former White House employee Elon Musk, has long said that the fate of his car company Tesla relies on robotaxis and repeated his claim to Bloomberg last month that he will debut his autonomous ride-hailing service in Austin, Texas, in June. (He said in 2019 that Tesla would be operating one million robotaxies by 2020).
The House proposal also comes as the first driverless 18-wheeler has hit the road. Aurora Innovation began operating its freight truck in Texas last month; it clocks up to 75 miles per hour, according to the company’s safety report.
“There seems to be this rush to implement and we still have questions about passenger vehicles in not even fully autonomous vehicles, but those that, at the moment of a crash, are on autopilot,” said the former official. “Those are concerning, those are alarming, and now we’re talking about 18-wheelers, big rigs, that are driving at higher speeds … We’re still grappling with the safety of passenger cars in autonomous modes and here we are rushing to implement the same or similar technology on semis.”
Multiple state attorneys general have objected to the AI regulation ban proposal, telling House and Senate leadership in a letter sent in May that the fallout from the provision would be vast. State laws protecting against AI-generated explicit material, deepfakes designed to mislead voters, spam calls and texts, and algorithms that jack up rent would all get the axe, they wrote. So would data privacy legislation enacted in 20 states.
“In the face of Congressional inaction on the emergence of real-world harms raised by the use of AI, states are likely to be the forum for addressing such issues. This bill would directly harm consumers, deprive them of rights currently held in many states, and prevent State AGs from fulfilling their mandate to protect consumers,” the attorneys general wrote.
States would also be unable to regulate the construction and operation of data centers, wrote Tim Storey, chief executive officer of the National Conference of State Legislatures, in a letter to House Energy and Commerce leaders. That would “severely limit” state and local officials’ ability to enforce zoning decisions, protect infrastructure such as power plants, and protect ratepayers against rate hikes.
Researchers said the proposal could trip over the Senate’s Byrd Rule, which bars “extraneous” matters unrelated to the budget from inclusion in a budget reconciliation bill. Blocking state-based regulations would not appear at first glance to have any federal budgetary effect, even if the action is vaguely intended to make the Commerce Department’s investment in AI systems more effective.
But scholars also noted that Byrd is not self-enforcing. Unless Senate Republicans preemptively take the provision out of their version of the bill, a senator would need to raise a point of order challenging the language to Senate Parliamentarian Elizabeth MacDonough. And even if she determined that the provision violated the rule, she is not the ultimate decision-maker. The Senate’s presiding officer on the day of the debate could simply ignore the parliamentarian, as senators did last month when they barreled ahead with a simple majority vote to eliminate an electric-vehicle mandate in California, which the Government Accountability Office said should have fallen under the 60-vote filibuster threshold. MacDonough backed the GAO’s determination.
“It’s unprecedented and a big deal to ignore the recommendation of the parliamentarian, but it’s certainly within the power of the presiding officer,” said Richard A. Arenberg, visiting professor of the practice of political science at Brown University, and the author of Congressional Procedure: A Practical Guide to the Legislative Process in the U.S. Congress. “It’s kind of a worn-out phrase but it would be the nuclear option … the minority would stop cooperating on anything, ending unanimous consent agreements and all sorts of things.”
The more typical procedure to preserve a provision that is challenged under the Byrd Rule is for the majority to gather 60 votes to overturn the ruling or to waive the Byrd Rule.
“If that provision is still in the bill, the burden is still on the majority to get 60 votes to keep it in,” said Sarah A. Binder, senior fellow at the Brookings Institution and a professor of political science specializing in Congress at George Washington University. “They absolutely could blow it up [by ignoring normal procedures], but we don’t know that they’re going to do that … it would not be following the order that comes along with the Budget Act.”
