This Autonomous Vehicle Firm is Being Investigated For Crashes With Safety Drivers At The Wheel

The act of driving is an extremely complex activity. So when companies set out to create a system that can do it in place of a real person, there were always going to be some teething pains. We’ve seen that time and time again in the form of crashes and traffic violations, even from the big names in the autonomous driving industry, including several from Waymo and Tesla.

Avride is another company that has been developing its own autonomous driving system for use in passenger vehicles and ride-sharing programs. Currently operating out of Austin, Texas, it’s been developing its product since 2017, first as a part of Yandex (the Russian equivalent of Google) before being spun off as its own company in 2020. While not nearly as widespread as Tesla or Waymo, Avride struck a deal with Uber to start offering autonomous rides using the company’s modified Hyundai Ioniq 5s (the deal also involves a tie-up between Uber Eats and Avride’s four-wheeled delivery robot).

Avride has been offering its autonomous driving services to the public via Uber since December 2025, and it doesn’t sound like it’s gone without issues. The National Highway Traffic Safety Administration (NHTSA) has opened an investigation into Avride for a series of crashes that allegedly occurred while the system was active, including an incident that allegedly caused an injury.

NHTSA says it’s reviewed video of each crash, though it hasn’t released those files to the public. Here’s what went down, according to the investigation’s preliminary findings:

The videos show instances of the AVs changing lanes into the path of or directly into other vehicles traveling in an adjacent lane and in close proximity to an AV; failing to slow or stop for slow-moving or stopped vehicles in the lane and path ahead; failing to slow for or avoid vehicles entering the lane and path ahead; and striking stationary objects partially obstructing the lane and path ahead.

According to the report, these crashes have caused property damage and, in one instance, an “alleged minor injury.” Though every instance of a collision allegedly occurred while the car was being operated by the autonomous driving system, both NHTSA and Avride say there was a trained safety operator on board, in the driver’s seat, at the time. From the sounds of it, it seems like this is one of those situations where the human monitor might’ve either not caught the mistake in time or simply wasn’t paying attention before a collision occurred (though that’s just speculation).

There’s no word on whether that minor injury was an Uber passenger, one of Avride’s trained operators, or someone outside the vehicle. I reached out to the company for some clarification, and here’s what an Avride representative told me:

Avride is aware of the Preliminary Evaluation opened by NHTSA. We welcome the opportunity to provide the agency with a deeper understanding of our safety protocols and technology. Transparency is a core value of our operations, and we look forward to a constructive dialogue that supports the continued progress of the autonomous vehicle industry.

We report incidents involving our vehicles to NHTSA—regardless of fault or severity—as part of our commitment to transparency and in accordance with applicable rules. In every reported incident, the vehicle was under the supervision of a trained safety operator on board. In most cases, the vehicle was traveling at low speeds and many of the events were precipitated by the actions of other road users.

That last part might sound like Avride is trying to downplay the incidents, but seeing as how only one alleged minor injury has been reported, I’m inclined to believe these cars aren’t sailing into stopped traffic at 30-plus mph. If I had to guess, it’s a lot of bumper taps, striking roadway debris, and other small fender-bender-type incidents. Avride may just be doing its due diligence by reporting every single incident, big or small.

NHTSA says these crashes “may indicate inappropriate assertiveness and insufficient competence” to operate on U.S. roadways, though, for what it’s worth, Avride says it’s actively addressing the problems that led to those incidents. Here’s what else the rep told me:

As part of our standard safety protocol, every incident undergoes a structured review that includes identifying and evaluating root causes. We have implemented targeted technical and operational mitigations to address our findings from each reported incident between December 2025 and March 2026, and have further enhanced overall system capabilities.

Our total operations have continued to grow, while the frequency of incidents relative to our mileage has steadily declined. We remain committed to transparency with NHTSA, and other key stakeholders, as we continue to safely scale our operations.

How the cars are getting into these crashes with a human supposedly monitoring the situation is still unclear. Avride doesn’t want to say whether human intervention could have stopped these incidents, though if they could have, then that’s a bad look on the supposed “training” these operators are getting, just as much as it’s the fault of the software itself. Either way, this is an instance where it sounds like the company is holding itself accountable, which, given how chaotic and wild-west the world of startup autonomous driving firms can feel at times, would be refreshing.

Top graphic image: Avride