New federal filings show five additional Tesla robotaxi crashes in Austin, as reporting scrutiny grows

Five newly logged incidents raise the Austin total to 14 since mid-2025 launch

Federal crash-report filings updated in early 2026 add five more reported collisions involving Tesla vehicles operating as part of the company’s supervised “robotaxi” program in Austin, Texas. The new entries bring the total number of reported Austin robotaxi crashes to 14 since the service began in June 2025.

The five incidents were submitted in January 2026 and describe crashes that occurred in December 2025 and January 2026. The filings list Model Y vehicles and indicate the automated driving system was “engaged” at the time of each crash. The records include basic descriptors such as vehicle movement and approximate speed, including low-speed collisions while backing and contact with other vehicles, including a bus and a heavy truck.

One earlier crash was revised to reflect a hospitalization

In addition to the five newly recorded crashes, an older Austin robotaxi crash entry from July 2025 was revised in a later report version to reflect an injury classified as “minor with hospitalization.” The July 2025 incident is described in the filing as a low-speed collision during a right turn involving an SUV. The updated injury classification indicates at least one person was transported for hospital treatment associated with that crash.

How these crashes appear in federal databases—and what is missing

The incidents are recorded under the National Highway Traffic Safety Administration’s Standing General Order (SGO) crash-reporting framework, which requires certain manufacturers and operators to report qualifying crashes involving automated driving systems (SAE Levels 3–5) and, separately, Level 2 driver-assistance systems when specific severity thresholds are met.

For the Austin robotaxi entries, key narrative fields are heavily redacted as confidential business information, limiting independent assessment of factors such as fault, road-user behavior, and whether a safety monitor intervened before impact. The SGO datasets also come with an important limitation: they are not normalized by miles traveled, fleet size, or operating conditions, making direct comparisons across operators difficult without additional exposure data.

Crash-rate comparisons hinge on mileage disclosures

Tesla has disclosed cumulative paid mileage figures for its robotaxi operations in corporate communications, which some analysts use to estimate a crash rate per mile for the Austin fleet. Using those disclosures as context, the current count of 14 reported crashes has been characterized as occurring more frequently than typical crash rates described in Tesla’s own safety reporting for human driving. Those comparisons remain sensitive to assumptions about miles traveled in Austin specifically and about how completely crashes are captured in any single reporting system.

Regulatory landscape in Texas is shifting again in 2026

Separate from the federal reporting regime, Texas enacted a commercial authorization framework for Level 4 and Level 5 automated vehicles in 2025. State agencies finalized implementing rules with a scheduled effective date in late February 2026, and the authorization requirement is set to become enforceable on May 28, 2026. The state program gives the Texas Department of Motor Vehicles authority to issue and revoke authorizations for commercial automated-vehicle operations and to restrict operations when public safety is at risk.

• Five additional Austin robotaxi crashes were logged in the latest set of federal filings, bringing the total to 14 since June 2025.
• A July 2025 crash entry was later revised to reflect a hospitalization-related injury.
• Federal crash datasets provide a structured reporting channel but remain limited by redactions and lack of mileage normalization.

The newly updated filings expand the public record of Austin robotaxi incidents while leaving key questions—such as detailed circumstances and system performance—unanswered due to redactions and incomplete exposure data.