TL;DR
A Reuters investigation found that 7 out of 9 Tesla data tags will not mount to FSD. They regularly saw the system speed up and fail on camera.
Reuters interviewed nine former Tesla data taggers and a former self-driving engineer about their thoughts on Tesla’s full self-driving mode. Seven out of nine data professionals said they would not ride in a FSD-powered Tesla. One said they wouldn’t ride a Tesla robot taxi.If you pay me.“
“We’ve all seen it fail,An insider told Reuters. A former self-employed engineer agreed: “Don’t trust Elon on this one at all.“They were referring to Musk’s statement that Tesla’s cars are ready.”safely unattended” rides.
The job of the data taggers was to study hours of FSD footage and train the car’s software to avoid past mistakes. They had direct access to terabytes of personal driving data. At least five people told Reuters they regularly saw clips of Teslas speeding while working at FSD.
Speeding was considered a low priority by engineers and managers. More attention was paid to extraneous problems such as unusual road configurations or rare lighting conditions. Priority was given to exceeding the normal speed limit, which affects every driver and every road.
💜 of EU technology
The latest rumblings from the EU tech scene, a story from our wise founder Boris and some questionable AI art. Free in your inbox every week. Register now!
The investigation comes as Tesla expands FSD availability to new markets. Tesla confirmed FSD availability in China last weekthough it’s unclear whether mainstream consumers will still be able to activate the system. The FSD (Supervised) system is classified as Level 2, requiring constant driver attention. A fully autonomous unsupervised version is being tested exclusively in the robotaxis fleet in Austin, Texas.
In recent months, there have been a number of incidents related to FSD. FSD-powered Teslas drove into lakes, off bridges, and into the path of oncoming trains. These are events that are on the agenda of the media. Testimonials from data taggers indicate that internal imaging has a larger catalog of failures.
The gap between Musk’s claims and the system’s performance remains a persistent problem. Musk has repeatedly promised fully autonomous driving since 2016. Every deadline has been passed without delivery. The company’s robot taxi service in Austin operates in a geo-fenced area where safety drivers are available remotely.
Waymo’s closure due to flooding this month demonstrated that even the most advanced autonomous control systems have failure modes under normal conditions. Tesla’s approach is fundamentally different from Waymo’s: a consumer car modified for camera perception and autonomy alone versus a multi-sensor combination and purpose-built robotics.
The term data tagger is important because these workers are the closest to raw performance data. They don’t see marketing materials or earnings projections. They watch hours of video showing how the software behaves on public roads. Seven out of nine people wouldn’t ride the product they helped build.
Tesla did not respond to Reuters’ request for comment. The company has previously stated that FSD (Supervised) requires active driver supervision and that its safety statistics show the system outperforms human drivers on a per-mile basis. A former engineer interviewed by Reuters disputed these statistics.
The investigation raises a question that Tesla’s regulatory filings and marketing haven’t answered: If the people who train the AI ​​don’t trust it, why should the people who ride it?
