As Tesla (TSLA) races to meet a March 9 deadline to deliver critical full-self driving (FSD) data to the National Highway Traffic Safety Administration (NHTSA), investors are once again scrutinizing the safety of its autonomous technology.
Since the company launched its unsupervised FSD robotaxi service in Austin in June 2025, 14 collisions involving the robotaxis have been reported. Headlines have screamed “accident-prone” and “safety myth shattered,” but a closer look at the numbers reveals a more nuanced picture—one where Tesla’s safety advantage may still be intact.
According to Tesla’s own safety report, vehicles using supervised FSD are involved in a major collision once every 5.3 million miles. In contrast, the average U.S. driver experiences a major crash every 660,000 miles. Even when accounting for all assisted-driving phases, that puts FSD at more than eight times safer than human drivers—a cornerstone of Tesla’s safety narrative.
The data Tesla is preparing to submit to NHTSA covers thousands of records from early FSD versions, though only 58 incidents were flagged as crashes when the investigation first opened. While the delayed submission has sparked speculation, the key lies in the timeline: the records span from FSD’s infancy through iterative updates. If the data shows a steady decline in incident rates with newer versions, it would underscore Tesla’s algorithmic progress. Conversely, if old flaws persist in current versions, investor confidence could waver.
The 14 robotaxi collisions—accumulated over roughly 800,000 miles—translate to a rate of about one crash every 57,000 miles. At first glance, that seems far worse than the U.S. average of one minor collision every 222,000 miles. But comparing the two is an apples-to-oranges exercise.
A closer look at the 14 incidents reveals telling patterns:
Elon Musk once noted, “All it takes is one in 10,000 trips to go wrong and you’ve got an issue.” Based on Waymo’s estimated average trip length of 4.3 miles, Tesla’s 800,000 robotaxi miles equate to about 186,000 trips. With 14 incidents, that’s roughly one collision per 13,300 trips—a rate that, while needing improvement, is far from “out of control.”
For investors, the March 9 submission offers two crucial insights: first, the violation rates and unresolved issues from early FSD versions; second, whether NHTSA plans further action based on historical data. If the data shows that early problems have been largely addressed in later updates, the market impact may be limited.
More important is the trajectory of robotaxi safety as the fleet scales. The 14 incidents so far have involved no serious injuries, and most were minor. As the number of robotaxis grows, absolute incidents will inevitably rise—but as long as the incident rate continues to decline and severe accidents remain rare, regulators and the public are likely to stay tolerant.
Tesla’s edge lies in its massive fleet and rapid iteration capability. Waymo, too, faced frequent complaints and crashes in its early San Francisco days, yet its per-mile collision rate has since fallen to acceptable levels. With real-world data pouring in from hundreds of thousands of vehicles, Tesla’s algorithm may evolve even faster.
The current data suggests that supervised FSD remains safer than human drivers, and robotaxi incidents—though numerous—are largely low-speed, stationary, or not the system’s fault, with a downward trend in frequency. If the March 9 report confirms that Tesla is systematically resolving early version issues, the answer to “how long can Tesla’s FSD safety card hold?” might be: at least until autonomous driving goes mainstream.
Still, risks remain. If newer FSD versions fail to fix old bugs, or if NHTSA deems certain flaws systemic and orders a recall, Tesla’s robotaxi ambitions could face a real setback. Only then would the safety card truly be tested.
