A House subcommittee approved critical legislation for the fast-developing driverless car movement this week that sets specific rules for how many such vehicles can be on U.S. roads and the federal government’s role in regulating them. While it would give companies developing the technology the national framework they want, it also raises questions over who is best-suited to ensure this cutting-edge technology is safe.

Under proposed rules that will come before the full House for a vote in September, manufacturers could put up to 100,000 autonomous vehicles per company on U.S. roads every year, with a requirement that they certify the safety of those vehicles with the National Highway Traffic Safety Administration, or NHTSA. If approved, the proposal would preempt the current patchwork of state-level rules for operating self-driving cars and trucks. They build off a basic set of guidelines issued by President Barack Obama’s Transportation Department last September.

“The need for this legislation was laid out by the Obama Administration,” said Rep. Bob Latta, the Ohio Republican who chairs the House Subcommittee on Digital Commerce and Consumer Protection. “From the front bumper to the back bumper, whether it is a car, a pickup truck or a van, how the vehicle works and is designed should be the province of the federal government as has been the case for more than 50 years.”

California, which has issued permits to 36 companies testing about 200 autonomous vehicles in the state, would no longer be able to set its own guidelines if the federal proposal becomes law. But whether federal safety officials ultimately are better-suited to validate manufacturers’ safety claims for the artificial intelligence, algorithms and next-generation vision and sensing technologies that go into self-driving cars remains in question.

“There’s this whole gray ground about who is in the best position to regulate these new cars, which go beyond the traditional car into almost driving a computer and software,” said attorney Sharon Klein, a partner in the autonomous vehicle practice for Pepper Hamilton LLP. “There’s been this debate about whether NHTSA as an agency should continue to be involved in deployment as well.”

The new rules rely on companies to certify the safety of their systems. The concern is that NHTSA, the country’s primary auto safety regulator, lacks the software and computer science expertise needed to validate their technical claims, Klein said.

“It comes down to a question of who is going to be watching the manufacturers?” she told Forbes.

Along with the growing number of robotic vehicles in California, states including Michigan, Massachusetts and Nevada are home to ever larger test fleets. Waymo, the Alphabet Inc. unit created last year to commercialize Google’s self-driving car technology, has already begun a large-scale public test in Arizona that may eventually grow to hundreds of driverless vehicles ferrying people around Phoenix.

General Motors, Uber, Lyft, Ford companies are following suit, and Tesla intends to give its electric vehicles full autonomous capability within the next year or two via over-the-air updates.

In 2016, NHTSA and the Transportation Department defined 15 areas that manufacturers need to comply with when they put self-driving vehicles on the road. That basic framework will put the agency in an administrative role, ensuring certain benchmarks are met, despite lacking specific knowledge of the software and algorithms powering autonomous systems.

“There’s no way NHTSA has the technical capability to do this right now,” said Mike Ramsey, an automotive tech analyst with Gartner Research in Detroit.

“NHTSA is really the arbiter of ‘does it appear that you have done all the things you need to do?’ It’s counting on the fear of deaths and lawsuits that could arise to prevent carmakers from doing something egregious,” he said.

Insurance companies will also act as a “shadow regulator” to keep manufacturers honest in their technical claims, Ramsey said. “The fact is unsafe (autonomous) technology would quickly become uninsurable.”

Consumer Watchdog, a safety advocacy group, called on congress to give NHTSA more resources to better address this major technological shift and opposed the reduced role of individual states to regulate robotic cars and trucks.

“Lost in the hyperbole over robot cars is a realistic assessment of the likely costs to both consumers and taxpayers particularly over the coming decades, when robot cars and human drivers will share a ‘hybrid highway,’ John Simpson, director of the Santa Monica, California-based group’s privacy project, said in a statement.

“Pre-empting the states’ ability to fill the void left by federal inaction leaves us at the mercy of manufacturers as they use our public highways as their private laboratories however they wish with no safety protections at all,” said Simpson, who called on California House members to oppose the subcommittee proposal.

For its part, the California DMV said it doesn’t comment on pending legislation. The department has been tasked with setting regulations for the testing and deployment of autonomous vehicles on state roadways since 2014.

Alan Ohnsman covers the intersection of technology, autos and mobility. Follow him on Twitter and LinkedIn