NHTSA opens probe into Avride crashes

Robotaxis are back in the federal hot seat. This time it’s Avride — the autonomous driving startup that runs rides on Uber in Dallas and has been expanding in Texas. On May 8, NHTSA opened a preliminary evaluation into Avride’s automated driving system after a cluster of crashes in Dallas and Austin raised questions about how the vehicles behave in ordinary traffic. The stakes are simple: if the software is making the wrong call in common road situations, that is not a weird edge case. That is the whole product. ### What did NHTSA actually do? NHTSA’s Office of Defects Investigation opened a Preliminary Evaluation, the first formal step in a federal defect probe. That stage is meant to figure out the scope and severity of a possible safety problem and decide whether to close the case, escalate it, or push toward a recall. NHTSA says a Preliminary Evaluation is generally worked on within about eight months. (static.nhtsa.gov) ### What is the agency worried about? The concern is not just “there were crashes.” NHTSA’s own summary points to specific driving behaviors — lane changes, responses to vehicles entering or already in the lane ahead, and crashes into stationary objects that were partially blocking the travel lane. In plain English, the agency is looking at whether the system is too aggressive, too limited, or both. Reuters’ account captured the same theme with NHTSA flagging possible “excessive assertiveness” and “insufficient capability.” (static.nhtsa.gov) ### How many crashes are we talking about? The federal case centers on 16 crashes in Dallas and Austin between December 2025 and March 2026. Some reports in circulation mention higher totals, but the clearest primary-source count tied to the opened probe is 16. NHTSA says the automated driving system was engaged in every crash it reviewed for this investigation, and at least one incident involved a minor injury alongside property damage. (static.nhtsa.gov) ### Were these fully driverless cars? Not quite. The vehicles under review had in-vehicle operators sitting in the driver’s seat while the automated driving system was engaged. That matters because this is not yet the cleanest version of the robotaxi promise — software still had human supervision available, and the crashes happened anyway. The question now is whether the human role was meaningful backup or mostly a legal and operational safety layer that could not reliably prevent bad decisions in time. (static.nhtsa.gov) That last part is an inference, but it follows from the setup NHTSA described. ### Why does Uber keep coming up? Because Avride is one of Uber’s robotaxi partners in Dallas. Uber has been building the marketplace layer for autonomous rides while relying on outside AV companies to supply the vehicles and driving stack. So even though the probe is about Avride’s system, it also lands on Uber’s broader strategy: scale robotaxi service through partnerships, then hope each partner clears the safety and regulatory bar. (static.nhtsa.gov) ### Why Texas? Texas has become one of the main proving grounds for commercial autonomous driving. The state has relatively permissive rules, big metro areas, and companies eager to launch paid service fast. That makes Texas useful for deployment — but also the place where weak spots show up first, because the systems are mixing with real traffic, delivery vehicles, construction zones, and impatient human drivers every day. The Avride probe is basically a stress test of that model. (wfaa.com) ### What happens next? NHTSA can demand more data, review technical safeguards, and decide whether the issue is narrow, systemic, or serious enough to escalate into a deeper engineering analysis or a recall path. So this is not a verdict yet. But it is more than a headline. A federal defect probe means Avride now has to show that these crashes were either limited, explainable, and fixable — or risk much heavier scrutiny. (wfaa.com) ### Bottom line The important shift is not just that Avride crashed. Robotaxis crash sometimes — the industry already knows that. The important shift is that federal regulators now seem focused on the style of the driving itself: whether the software behaves like a cautious road user or an overconfident one. For any AV company trying to scale in Texas, that is the test that now matters most. (static.nhtsa.gov)