Tesla Releases Unredacted Reports for 19 Robotaxi Incidents

Tesla has updated its historical filings with unredacted reports for 19 incidents occurring within its test fleet in Austin, Texas. The automaker had stood alone for the better part of a year as the only autonomous vehicle operator filing safety data that hid contextual specifics of its accidents. Every single incident report previously featured the same boilerplate phrase: [REDACTED, MAY CONTAIN CONFIDENTIAL BUSINESS INFORMATION]. This created a longstanding point of tension with safety researchers and public advocates.

The newly visible information comes directly from data published on the National Highway Traffic Safety Administration (NHTSA) website. It falls under a Standing General Order that requires companies testing or deploying Automated Driving Systems to maintain a transparent log of public road accidents. The unredacted summaries show that the vast majority of these incidents were minor, low-speed events that were not the fault of Tesla’s Full Self-Driving software.

A substantial portion of the 19 reported incidents involves manually driven cars crashing into stationary Teslas. Multiple summaries describe the car being at a complete halt at an intersection or red light when a human driver behind them slowly rolled or crept forward, rear-ending the vehicle. In dense, low-speed urban environments, external actors caused several no-fault incidents, such as a pedicab clipping the right side mirror of a stopped Tesla or a motor scooter steering right into the rear bumper before hopping a curb to escape onto the sidewalk.

Out of Tesla’s 19 total reported incidents, an overwhelming majority resulted in zero injuries. The data shows only two minor injuries across the entire set, with absolutely no high-speed impacts, severe structural failures, or critical injury crashes. Crucially, every single reported event occurred while human safety monitors were present in the front seats. None of the entries stems from Tesla’s fully unsupervised operations, and none suggests a fundamental failure of the core machine-learning logic.

The autonomous vehicle’s own minor errors were restricted to tight parking lots or alleyways, such as a tire scraping a curb or a side mirror striking a metal chain fence or utility pole while reversing at single-digit speeds. This pattern of human drivers being responsible for the vast majority of autonomous fleet collisions is becoming a standard benchmark for the self-driving industry. The trend was recently mirrored when one of the newly mass-produced Cybercabs that Tesla is currently testing on public roads before adding them to the Robotaxi network was rear-ended by a manually driven vehicle.

Tesla is gradually transitioning its Robotaxi service to fully autonomous operation, with no safety driver or even monitor onboard. The company’s recent expansion into Dallas and Houston has been entirely unsupervised. Competitors like Waymo and Zoox have logged higher total incident counts, but they also operate much larger fleets. By providing full transparency to the public, Tesla effectively shifts the conversation back to its safety advantages.