The federal auto safety agency's proposed rewrite of FMVSS No. 135, the federal light vehicle brake standard, would drop the foot pedal requirement for driverless vehicles.
For decades, the brake pedal was the regulatory proof that a human was still in charge of a passenger car. On June 26, 2026, NHTSA published a proposed rule in the Federal Register that, on its face, just removes that pedal for driverless vehicles. The face is misleading. The pedal was never just a piece of hardware. Under the existing rule, it was the regulatory stand-in for a human reflex the agency could test, label, and recall. The proposed rule quietly drops the stand-in. The change being negotiated is not 'kill the brake pedal.' It is 'stop treating the human body as the safety-assurance unit.'
The standard in question is FMVSS No. 135, the federal light-vehicle brake-systems regulation that has anchored occupant safety to a pedal-and-booster architecture for decades. Its core assumption was mechanical and human at once: a driver sees a hazard, a foot pushes a pedal, hydraulic pressure stops the car. Every brake-system test, label, and failure mode in the standard was built around that chain. NHTSA's press release makes the agency's logic explicit: in a vehicle designed to operate with no human in the control seat, the requirement to install a foot brake 'impedes innovation' because there is no occupant positioned to actuate it. The standard's old unit-of-account no longer matches the new design.
What the proposed rule actually does is narrow. Per Insurance Journal's read of the docket and The Register's summary, the pedal removal applies only to vehicles whose Automated Driving System is the sole means of control inside a defined operational design domain. Manually driven cars and dual-mode vehicles keep the existing FMVSS No. 135 pedal architecture unchanged. The framing matters: this is not a deregulatory shotgun blast at light-vehicle safety. It is a scoped carve-out aimed at the small but growing fleet of vehicles where no human foot is available, or expected, to intervene.
The deeper shift is conceptual. For decades, FMVSS No. 135's pedal requirement functioned as a regulator's guarantee that at least one safety-critical action in a passenger car was routed through a human reflex the agency could test. Removing the pedal for ADS-equipped vehicles means the standard no longer assumes that guarantee exists. Instead, the assurance layer has to come from somewhere else: the ADS software stack, the redundant braking architecture, the operational design domain limits, the post-deployment reporting regime, and whatever oversight NHTSA builds around them. The pedal was the proxy. With the proxy gone, the regulatory question is whether the replacement stack is strong enough to bear the weight.
That is the part the headlines skip. The Register frames the proposal as part of the Transportation Department's broader 'commonsense' AV reform push, which signals the political posture: less hardware mandate, more deference to operator safety cases. Insurance Journal treats it as a scoped technical change. Neither piece yet answers the central question: when the pedal disappears, what specifically in the ADS safety case substitutes for it inside FMVSS No. 135? The NPRM text will tell us, but the trade-press summaries do not.
The exact comment-window length and effective-date mechanics are not yet pinned down in secondary coverage and should be read from the Federal Register docket itself. What to watch next is narrower than the headlines suggest. First, the docket's exact delineation of which vehicle classes qualify as 'ADS-equipped' and whether dual-mode fallbacks are grandfathered. Second, how NHTSA handles brake-system redundancy tests for driverless platforms where no human foot can serve as the final safety net. Third, whether safety advocates treat the human-fallback removal as a rollback or a reasonable evolution. Each of those turns on the same underlying question the pedal used to answer silently: who vouches for the car when the driver is not there?