Tesla's widely promoted claim that its Full Self-Driving (FSD) technology is 10 times safer than human drivers rests on a fundamental comparison error that inflates the safety advantage by roughly three times, according to a detailed Reuters investigation. The automaker counted Tesla crashes involving airbag deployments and compared them against federal data on all crashes requiring tow trucks, a far less restrictive standard. This apples-and-oranges approach produced inflated safety statistics that Tesla executives, including Chief Financial Officer Taneja, have used to promote the technology as a personal chauffeur alternative at $99 per month.

What's Wrong With Tesla's Safety Methodology?

The core problem stems from how Tesla selected its comparison groups. When researchers at the University of Michigan Transportation Research Institute reanalyzed the data using consistent criteria, they found a significantly different picture. Using airbag-involved crashes for both Tesla vehicles and other cars, the analysis showed Teslas using FSD or Autopilot travel about three times farther between crashes where airbags deployed, not 10 times.

Marco Benedetti, an assistant research scientist at the University of Michigan Transportation Research Institute and former National Highway Traffic Safety Administration (NHTSA) statistician, performed the analysis. Two other traffic-safety researchers independently verified his calculations and agreed with the findings. Yet even this three-times figure may overstate FSD's safety advantage because of several other methodological flaws in Tesla's approach.

Tesla's methodology contains multiple additional problems that cast doubt on whether FSD is actually safer at all:

• Vehicle Age Bias: Tesla compares its vehicles, which average 4.1 years old, against all U.S. vehicles averaging 12.8 years old. Newer cars across all manufacturers have access to modern safety features like blind-spot monitoring and automatic emergency braking, making the comparison unfair.
• Narrow Crash Window: Tesla only counts crashes occurring with FSD switched on or within five seconds of deactivation, while the U.S. government requires reporting crashes within 30 seconds of advanced driver-assistance system deactivation.
• Selective Data Inclusion: Tesla initially included Autopilot data alongside FSD, inflating miles-between-crashes figures because Autopilot is less sophisticated and primarily used on highways where crashes occur less frequently than on city streets.

How Does Waymo's Approach Differ?

Alphabet's Waymo takes a markedly different approach to safety claims. The company operates fully driverless robotaxis in 11 U.S. metropolitan areas and compares its vehicles directly to human-driven vehicles in similar conditions and neighborhoods. Waymo examines specific crash rates, such as those involving airbag deployments or serious injuries, for both its robotaxis and human-driven cars operating in the same markets.

"We've got to be really careful with the language we use. You need very specific research questions and very specific conclusions," said John Scanlon, a Waymo safety researcher.

John Scanlon, Safety Researcher at Waymo

Waymo also publishes its safety statistics in peer-reviewed journals and collaborates with outside researchers, openly acknowledging shortcomings in its data. Tesla, by contrast, keeps its underlying crash data secret and publishes only top-line claims without seeking peer review.

What Are Tesla Employees Actually Seeing?

Inside Tesla, data labelers who train FSD have witnessed a different reality than what executives promote publicly. Three former employees described videos showing Teslas failing to recognize animals and striking them at speed without braking. Five former employees reported that specific teams focused on FSD's problems recognizing school buses, a concern that safety advocates have highlighted in Super Bowl advertisements.

The data-labeling environment itself reveals organizational strain. Five former data labelers described a disjointed work environment with shifting priorities based on directives from Elon Musk and FSD engineers. Chronic turnover plagued the unit due to monotonous work and low pay. Tesla leadership often launched new projects reactively, responding to social media videos or news reports showing FSD failures, such as efforts to address how sunlight could blind exterior cameras or how the system failed at railroad crossings.

Speeding emerged as a recurring issue in FSD videos reviewed by employees. Five former employees said FSD clips regularly showed vehicles exceeding speed limits, which engineers treated as a low-priority problem. One employee reported seeing Teslas regularly exceeding speed limits by 20 to 30 miles per hour after Tesla introduced an FSD "Mad Max" mode enabling more-aggressive driving. Another labeler witnessed an FSD-piloted vehicle traveling 60 miles per hour in a 25-mile-per-hour zone.

Steps to Evaluate Self-Driving Safety Claims

When comparing safety statistics from autonomous vehicle companies, consumers and regulators should apply rigorous standards to ensure claims are credible and comparable:

• Check Comparison Consistency: Verify that crash definitions are identical across both the autonomous vehicle fleet and the human driver baseline. Airbag deployments should be the standard metric for both groups, not different thresholds like tow-truck removals.
• Examine Vehicle Age Matching: Ensure the autonomous vehicles are compared against human-driven vehicles of similar age, since newer cars across all manufacturers have better safety features that inflate their crash resistance.
• Look for Peer Review: Safety claims published in peer-reviewed journals with outside researcher collaboration carry more credibility than internal company statistics kept confidential.
• Assess Operating Conditions: Confirm that comparisons account for the types of roads and neighborhoods where autonomous vehicles operate, since highway driving differs significantly from urban environments.

Tesla's claims about FSD's potential to save over 32,000 lives and prevent more than 1.9 million injuries annually rest on unrealistic assumptions, according to traffic-safety researchers. These figures assume every U.S. vehicle, including freight trucks and motorcycles, would be replaced by an FSD-enabled Tesla that is at least seven times safer than its predecessor. Such projections lack grounding in real-world deployment data.

The fundamental issue, according to 10 traffic-safety researchers cited in the investigation, is that FSD is not a truly autonomous system. Tesla compares average human drivers to other average human drivers who happen to be using FSD. The company fails to account for the fact that drivers can turn FSD on and off, and research shows motorists often avoid using advanced driver-assistance systems in complicated traffic situations where the technology feels unsafe. Tesla's own data shows FSD is used mostly on highways, where driving is inherently simpler and safer.

" }