Elon Musk says Tesla is currently running about 500 autonomous Model Y robotaxis across Austin and San Francisco. That’s still a relatively small footprint compared with larger self-driving ride-hailing operations in the U.S., but Tesla’s fleet is quickly drawing attention for another reason: the pace at which incidents are being reported as the service shifts deeper into fully unsupervised operation.
These robotaxis aren’t special prototype vehicles. They’re described as off-the-shelf 2026 Tesla Model Y units running a closed version of Tesla’s Full Self-Driving (FSD) software designed to operate without a human safety monitor. And according to the latest self-driving crash incident report Tesla is required to file with the National Highway Traffic Safety Administration (NHTSA), the Austin robotaxi fleet logged five incidents in a single month.
That monthly figure stands out when compared with last year’s total of nine incidents in Austin, around the time Tesla’s robotaxi platform began operating there. What makes the comparison especially important is timing: Tesla reportedly began allowing passengers to ride without a human safety monitor in January. Whether the step toward completely unsupervised rides is directly tied to the rising incident count isn’t confirmed, because Tesla has also been scaling the size of its ride-share fleet. More vehicles on the road can naturally lead to more reported events.
Still, the reported incident frequency for vehicles running unsupervised FSD comes out to roughly one incident every 57,000 miles. That’s a notable contrast to Tesla’s broader safety messaging around its driver-assist systems, including claims that its technology can be significantly safer than human drivers. Based on the numbers cited in the report content, the robotaxi incident rate appears far worse than a commonly cited human average of about one incident every 500,000 miles.
A major challenge in evaluating these comparisons is that Tesla does not publicly break down autonomous miles or incident data by software category, version, or operating mode. That lack of granularity makes it harder to assess how “Autopilot” statistics relate to an unsupervised robotaxi service. Driving with assistance on a mostly empty highway is a very different risk profile than navigating busy downtown streets during rush hour with no safety driver ready to intervene.
Tesla has reportedly categorized most robotaxi accidents as resulting in property damage, including low-speed incidents such as backing up in parking lots. However, the report content also notes that hospitalizations have occurred, underscoring why regulators and the public are watching these deployments closely.
Looking ahead, Tesla is preparing for a significant expansion of its robotaxi ride-share fleet into additional U.S. cities later this year. That ramp-up is expected to include the upcoming Cybercab, a two-seat vehicle described as having no pedals and no steering wheel. As Tesla scales beyond Austin and San Francisco and introduces purpose-built autonomous vehicles, the key question will be whether incident rates fall with software improvements—or rise simply because more driverless miles are being logged in more complex environments.
