Key Takeaways
- NHTSA’s June 26 NPRM removes the 1995-era requirement for hand- or foot-operated brake controls in vehicles designed exclusively for automated driving systems. The first time the federal brake standard has been updated for vehicles with no human driver.
- If you carry a standard personal auto policy today, this rulemaking does not change your premium. What it does is signal that purpose-built robotaxis and autonomous shuttles are closer to commercial deployment, and the industry’s current liability-assignment framework, which assumes a human driver, is not ready for that world.
- NHTSA simultaneously withdrew the AV STEP program on June 26, removing the voluntary transparency and reporting framework that would have given regulators visibility into ADS behavior in exchange for FMVSS exemptions, a move that leaves a meaningful oversight gap as commercial AV deployment accelerates.
What NHTSA Actually Filed on June 26
On June 26, 2026, the National Highway Traffic Safety Administration published a Notice of Proposed Rulemaking in the Federal Register proposing to amend Federal Motor Vehicle Safety Standard No. 135, which has governed light vehicle brake systems since 1995. The proposal removes the requirement for hand- or foot-operated brake controls in vehicles designed never to be operated by a human driver. Stopping distance performance requirements stay in place. Every other braking rule stays intact for any vehicle that retains a steering wheel or pedal. The comment period closes July 27, 2026.
That is a narrow change in the regulatory text. Its implications for the auto insurance market are not narrow at all.
On the same day, NHTSA formally withdrew its January 15, 2026 NPRM proposing the Automated Driving System-Equipped Vehicle Safety, Transparency, and Evaluation Program, known as AV STEP. That program would have required participating AV manufacturers and fleet operators to disclose safety data in exchange for potential FMVSS exemptions. Gone. What’s left is a rulemaking that clears a key hardware barrier for purpose-built robotaxis, combined with the removal of the one framework that would have given regulators real-time visibility into how those vehicles are actually performing on public roads.
The Insurance Framework That Does Not Exist Yet
Here is the problem that nobody in the NHTSA press release addresses: the auto insurance rating framework that every carrier in every state currently uses was built for a world where a human being is driving the vehicle. Every rate-determining variable in a standard personal auto filing, the driver’s age, driving history, credit score, telematics score, annual mileage, assumes a human is making driving decisions. The moment a vehicle is controlled entirely by an automated driving system, every one of those variables becomes irrelevant as a loss predictor.
I spent nine years writing auto policies at an independent agency and a captive carrier. I can tell you exactly how carriers currently handle autonomous vehicle exposure: they don’t. A customer who takes delivery of a vehicle with Level 2 driver assistance, like adaptive cruise control and lane-keeping, is rated the same way as someone driving a ten-year-old sedan with no technology at all. The SERFF rate filings for standard personal auto do not break out an AV component. The underwriting guidelines do not differentiate between a vehicle that requires constant human engagement and one that can manage highway driving without any input. That worked as long as no consumer-facing vehicle actually drove itself. It stops working the moment it does.
Waymo and Zoox operate commercially today, but under fleet structures that sidestep the personal auto market entirely. When AVs reach the personal auto space at any scale, carriers will need to answer a question the NAIC has not yet resolved: who is the insured driver when there is no driver? The liability assignment shifts from the human to the manufacturer and, potentially, to the software stack. That is a product liability question, not a personal auto question. The rating model is different. The reserve methodology is different. The reinsurance structure is different. None of that infrastructure exists for the personal auto market today.
NHTSA Administrator Jonathan Morrison framed the brake pedal rulemaking as tearing down pointless barriers under the Department of Transportation’s AV Framework. He is right that a physical pedal serves no purpose in a vehicle no human will ever operate. He is not wrong that legacy regulatory requirements were creating unworkable barriers for companies like Zoox, whose Amazon-backed robotaxi already holds an FMVSS exemption to operate in limited commercial settings. But the rulemaking accelerates the deployment timeline without adding any new insurance framework to match it. NHTSA acknowledged in the rulemaking itself that it is separately developing broader ADS safety performance standards for how autonomous vehicles behave on public roads. Those standards do not yet exist.
What This Means for Drivers Shopping Coverage Today
If you are renewing your policy this month, this rulemaking affects you in approximately zero ways. No carrier has filed a rate adjustment tied to FMVSS No. 135. No SERFF filing anywhere references AV liability as a component of standard personal auto pricing. Your premium is driven by your driving record, your vehicle’s physical damage history, local catastrophe model exposure, and severity trends in your state. None of which change because of a proposed update to a 1995 brake standard.
The story is about where rates go from here. NHTSA has now proposed five FMVSS updates in rapid succession as part of the AV Framework, clearing hardware requirements for vehicles with no human controls. Each one moves purpose-built AVs closer to commercial deployment at scale. When that deployment reaches the personal auto market, carriers will file rate adjustments, and those filings will be messy, because the actuarial data does not yet exist to price AV liability cleanly. The first few years of AV personal auto pricing will look a lot like the first few years of electric vehicle pricing: carriers guessing at loss costs with limited claims experience, some getting it wrong in both directions, and policyholders in the early-adopter cohort paying for the uncertainty.
If you are currently shopping car insurance rates, the practical takeaway is simpler. Driver-assistance technology in your current vehicle does not change your personal auto rating category unless you are enrolled in a telematics program like Progressive’s Snapshot or State Farm’s Drive Safe and Save. The vehicles affected by today’s rulemaking are purpose-built robotaxis with no pedals, not the vehicles your neighbors are driving. You will not see a line item for AV liability on your renewal letter anytime soon.
What you should watch is the NAIC. The National Association of Insurance Commissioners has been developing a framework for third-party data and predictive models, including how ADS-generated driving data could eventually be used in underwriting. That framework is still in draft. If and when NHTSA finalizes the FMVSS No. 135 changes and purpose-built AVs begin commercial personal deployment, the NAIC model law work becomes the next domino. The public comment deadline for NHTSA’s proposal is July 27, 2026, and the insurance industry’s silence in that comment period would be its own kind of answer about how ready the market is for what comes next.
The AV STEP withdrawal is the detail that deserves more attention than it is getting. That program would have required manufacturers to publicly disclose safety performance data as a condition of receiving FMVSS flexibility. Its removal means AV operators get regulatory runway without the transparency obligation. For insurers trying to build AV rating models, that missing data stream is a real problem. You cannot price a risk you cannot measure. NHTSA says it will use its defect enforcement authority to oversee unsafe ADS behavior in the interim. Defect enforcement is reactive. Insurance pricing needs to be prospective. Those two things are not the same.
