After years of cautious permitting, US and global regulators in 2025 and 2026 cleared robotaxis to charge fares without steering wheels, expanded Waymo to ten cities, and launched federal AV safety rulemaking, while IIHS data showed Waymo crashing 68 percent less than human drivers. Yet the same period produced a mass Baidu robotaxi freeze in Wuhan, a three-month Chinese permit suspension, a California court ruling against Tesla FSD marketing, and a federal order forcing all operators to fix emergency-response failures, revealing that the regulatory green light is conditional and contested.
The autonomous vehicle industry crossed a threshold in 2025 and 2026 that years of cautious permitting had kept just out of reach. Regulators in the United States, China, and Europe moved from experimental tolerance to active commercial authorization, clearing robotaxis to charge fares, operate on freeways, and run without steering wheels or pedals. The pace of approvals accelerated faster than the frameworks governing liability, emergency response, and fleet-wide failure modes, producing a regulatory environment that is simultaneously more permissive and more contested than at any prior point.
The pattern matters now because the decisions made in this window will set the baseline for what autonomous vehicles are legally permitted to do at scale. NHTSA is rewriting the physical standards that define what counts as a car. China just reopened its market after a three-month freeze triggered by a mass fleet failure. The UN has adopted its first global ADS framework. Independent safety data is arriving for the first time. Each of these developments locks in assumptions that will be difficult to reverse once fleets number in the tens of thousands.
The Commercial Expansion Wave
The opening move came from Amazon's autonomous vehicle unit. On September 10, 2025, Zoox Opens First Public Robotaxi Rides in Las Vegas, deploying a purpose-built vehicle with no driver controls or steering wheel to the general public, the first such deployment in US history. Two months later, Zoox extended the experiment to California, with Amazon's Zoox Launches Public Robotaxi Service in San Francisco offering free rides in parts of the city.
Waymo moved in parallel on the operational side. On November 12, 2025, Waymo Robotaxis Hit Freeways in SF, LA, and Phoenix, the first time fully driverless ride-hail vehicles operated on interstates at commercial scale, cutting some trip times by up to 50 percent. Ten days later, California DMV Clears Waymo for Statewide Expansion, granting authorization across the entire Bay Area, Sacramento, and nearly all of Southern California down to the Mexican border, the broadest single regulatory expansion of a commercial robotaxi service in US history.
By February 2026, Waymo Robotaxis Live in 10 U.S. Markets, adding Dallas, Houston, San Antonio, and Orlando. By late March, Waymo hits 500,000 paid robotaxi rides per week across 10 US cities, a 10x increase from 50,000 weekly rides in under two years. Tesla joined the commercial operators on January 22, 2026, when Tesla Launches First Robotaxi Rides Without Safety Monitors, years after Elon Musk's original 2020 target. Tesla then expanded to Houston and Dallas in April 2026 with Tesla Launches Robotaxi Service in Houston and Dallas, and moved into Miami in July with Tesla Launches Fully Driverless Robotaxi in Miami, its first driverless deployment in a new city without a safety-monitor phase.
The international dimension arrived in April 2026, when Zagreb becomes first European city with public robotaxi service, with Verne, Pony.ai, and Uber launching an autonomous robotaxi service on public streets in Croatia. Verne, a Croatian startup, owns the fleet and manages operations, Pony.ai provides the autonomous driving technology, and Uber integrates the service into its platform. In June, UN adopts first global regulations for autonomous driving systems, replacing fragmented national approaches with a unified framework for Level 4 deployment.
The Regulatory Architecture Catching Up
The commercial wave forced federal regulators to confront a structural problem: the existing rules were written for vehicles with steering wheels and pedals, and the 2,500-unit annual production cap was already constraining operators. On July 26, 2026, NHTSA proposed to resolve both issues at once. The NHTSA Proposes Rule to Remove Pedal Mandate for AVs would eliminate the steering-wheel and pedal requirement entirely and create the first permanent federal certification pathway for pedal-free robotaxis, ending the production cap.
Four days later, the agency moved from proposal to action. On July 29, NHTSA issues new AV safety standards and Zoox exemption published new automated vehicle performance standards and granted Zoox a commercial deployment exemption. On July 31, Zoox becomes first steering-wheel-free robotaxi approved to charge for rides in US formalized the first-ever federal approval for paid rides in a vehicle without human controls. The boxy four-inward-seat pods, which had been offering free rides in Las Vegas and San Francisco, can now begin charging fares once state and local approvals are secured, subject to a 2,500-unit annual cap under the temporary exemption structure.
The safety data underpinning these approvals arrived on July 24, 2026, when IIHS: Waymo robotaxis crash 68% less than human drivers published the first major independent validation of a commercial robotaxi fleet at scale. The 68 percent reduction in police-reportable crashes is the first number regulators can cite that is not self-reported by an operator, and it materially changes the political calculus for further approvals.
Where the System Broke Down
The most consequential single event in this period was not an approval but a failure. On March 31, 2026, a cloud and dispatch system failure simultaneously halted 100+ Baidu Apollo Go Robotaxis Freeze in Wuhan Traffic, stranding passengers in live highway traffic for up to two hours. This was China's first reported mass shutdown of an autonomous vehicle fleet at scale, and the cause was not a sensor failure or a collision but a software dependency that took down more than 100 vehicles simultaneously.
The regulatory response was swift and sweeping. On April 29, 2026, China Suspends New AV Permits After Baidu Robotaxi Mass Outage halted issuance of all new Level 4 autonomous vehicle licenses, blocking fleet expansion, new pilots, and entry into additional cities. The freeze lasted three months. On July 23, 2026, China resumes robotaxi permits after 3-month freeze, with Momenta and Baidu among the first recipients, reopening the world's largest AV market to new deployments. The episode demonstrated that a single systemic failure can trigger a nationwide regulatory halt, and that the mechanism for that halt can be reversed, but not quickly.
The US had its own version of this dynamic. On July 13, 2026, US federal regulator orders all robotaxi operators to fix emergency-response failures, issuing a formal ultimatum to the entire autonomous vehicle industry demanding fixes for robotaxis that interfere with first responders by month's end. The order applied to all operators, not just one company, signaling that emergency-response interaction is a sector-wide gap rather than an isolated incident.
Tesla's problems were different in kind but equally concrete. On October 9, 2025, NHTSA Opens Probe Into 2.88M Tesla FSD Vehicles, citing more than 50 reports of traffic-safety violations including red-light running, wrong-way driving, and crashes causing injuries. On December 17, 2025, a CA judge rules Tesla FSD marketing false, orders 60-day fix, finding Tesla engaged in false advertising with its Autopilot and Full Self-Driving branding and giving the company 60 days to correct its marketing or face a 30-day suspension of its California manufacturing and sales license. The ruling drew a legal line between marketing language and demonstrated capability that other operators will need to navigate.
What to Watch
The NHTSA pedal-free rulemaking is the single most consequential regulatory action in the pipeline. If finalized, it removes the production cap and creates a permanent pathway for purpose-built robotaxis. The public comment period closed July 27, 2026, and the final rule will define the federal baseline for the next decade.
Waymo's stated goal of one million weekly rides by year-end 2026 is a commercial and regulatory stress test. At 500,000 weekly rides across 10 cities as of March 2026, reaching that target requires either doubling density in existing markets or accelerating the expansion announced in Las Vegas, Denver, San Diego, and Tampa in July 2026.
China's permit resumption terms will determine whether the three-month freeze produced structural changes to fleet management requirements or simply a pause. Baidu's cloud architecture was the proximate cause of the Wuhan outage; whether regulators imposed redundancy mandates as a condition of resumption is the key variable.
The emergency-response compliance deadline set by the July 13 federal order is a near-term forcing function. If operators cannot demonstrate fixes by month's end, the order creates a precedent for operational suspensions that would apply across the entire commercial robotaxi sector simultaneously.
The IIHS 68 percent crash-reduction figure will be tested against Tesla's expanding driverless footprint. Waymo's data covers millions of rides on a mature sensor stack. Tesla's FSD architecture is different, its NHTSA probe is active, and its marketing has been ruled false by a California court. Whether independent safety researchers produce comparable data for Tesla's driverless deployments in Houston, Dallas, and Miami will be the most watched empirical question in the sector over the next 12 months.
This piece was originally published on Present of AI, where we cover what AI is actually doing in the world, no hype. Read more or get it in your inbox.
Top comments (0)