Tesla's own AI trainers don't trust 'Full Self-Driving' or its safety stats, Reuters finds | Electrek

Skip to main content

Toggle social menu

Toggle dark mode

Search for:

Submit

Toggle search form

Toggle dark mode

Tesla

Tesla’s own AI trainers don’t trust ‘Full Self-Driving’ or its safety stats, Reuters finds

Fred Lambert | May 28 2026 - 6:33 am PT

180 Comments

A major Reuters investigation published today reveals that Tesla’s widely touted “Full Self-Driving” safety statistics are built on deeply flawed methodology — and that the company’s own data labelers, the workers who train the AI system, don’t trust the technology to drive them.

The report, based on interviews with nine former Tesla data labelers, a former self-driving engineer, and 11 traffic-safety researchers, paints a damning picture of the gap between Tesla’s safety marketing and the reality of its autonomous driving program.

Tesla’s safety stats inflated by a factor of 3

We’ve been calling out Tesla’s misleading FSD safety claims for a while now, and the Reuters investigation confirms the core problem with hard data.

Tesla CEO Elon Musk and other executives have repeatedly claimed that “Full Self-Driving” is up to 10 times safer than human drivers. Tesla CFO Vaibhav Taneja first made this claim last July, and Tesla Board Chair Robyn Denholm repeated it at a November shareholder meeting. Musk himself displayed a chart at that meeting claiming “85% less crashes.”<br>Advertisement - scroll for more content

Reuters found that a central comparison error inflated Tesla’s claimed safety level by a factor of three. Tesla counted crashes where airbags deployed in its own vehicles, then compared that number to federal data that includes all crashes requiring a tow truck — a far less severe threshold. Tow-truck crashes often don’t involve airbag deployments at all.

The critical point: the federal data Tesla used already included airbag-deployment crashes as a separate category. Tesla could have made a valid apples-to-apples comparison but chose not to.

When University of Michigan researcher Marco Benedetti performed the correct comparison — airbag crashes for Teslas versus airbag crashes for all vehicles — the result dropped from “10 times safer” to roughly three times farther between crashes. And even that figure is unreliable because of additional methodological problems, including the massive age gap between Tesla’s fleet (4.1 years average) and the overall U.S. fleet (12.8 years).

As Carnegie Mellon professor Phil Koopman put it: “It’s like saying: ‘My jet airplane is faster than your World War II bomber.’ Yeah, so, what’s your point?”

Ten of the 11 traffic-safety researchers who reviewed Tesla’s methodology for Reuters said the statistics amounted to misleading marketing rather than a serious safety investigation.

‘Don’t trust Elon on this’

Beyond the statistics, the Reuters report reveals what Tesla employees actually think about the technology they help build.

Seven of the nine former data labelers told Reuters they wouldn’t trust FSD to drive them. One said he wouldn’t ride in a Tesla robotaxi “if you fucking paid me.” A veteran self-driving engineer who reviewed Tesla crash data for years called the company’s safety claims “bullshit” and said: “Definitely, don’t trust Elon on this.”

The data labelers, based primarily in a Utah office, review video footage from the eight exterior cameras on Tesla vehicles using FSD. They described regularly seeing FSD fail at basic tasks: pulling over for emergency vehicles, giving motorcyclists enough space, braking on freeway off-ramps, and avoiding construction zones. In one incident, a Tesla drove into a construction zone and nearly struck workers.

A specialized team in Palo Alto, known internally as the “trauma team,” focused specifically on near-misses with pedestrians. Former employees described seeing clips of FSD-piloted Teslas nearly hitting children and failing to recognize pedestrians in crosswalks.

The report also details FSD regularly exceeding speed limits by 20 to 30 mph after Tesla introduced a “Mad Max” mode for more aggressive driving, with one labeler reporting an FSD vehicle traveling 60 mph in a 25-mph zone.

The robotaxi mapping that undermines Musk’s key claim

One of the most significant findings in the Reuters investigation is how Tesla extensively mapped its robotaxi operating zones before public launches — directly contradicting Musk’s central claim that Tesla’s approach doesn’t require the “laborious local mapping” used by rivals like Waymo.

For weeks before the October 2024 Cybercab unveiling at the Warner Bros. studio lot, staff tested prototypes every night from 6 p.m. until dawn, collecting video of the exact routes the cars would follow. Data labelers spent hundreds of hours annotating curbs and road markings to prevent embarrassing incidents.

The same thing happened before the Austin robotaxi launch in June 2025. Tesla extensively filmed features in...