People who trained Tesla’s self-driving software are warning FSD owners not to trust it.
A Reuters investigation has revealed hundreds of Tesla workers — known as data labellers — spend their days scrutinising footage captured by vehicles running the company’s Full Self-Driving (FSD) feature.
And what they’ve seen has raised serious concerns.
MORE: One bad June could rewrite Aus car history
According to nine former labellers and a former Tesla self-driving engineer, FSD has continued to struggle with basic manoeuvres in recent months.
Some video footage captured from Tesla’s FSD cameras show cars striking cats, dogs, and deer, and in some cases, near-misses with children playing in the street.
As a result, seven of the former data labellers and a former Tesla self-driving engineer told Reuters they wouldn’t trust revolutionary self-driving technology to drive them.
One former labeller said simply: “We have all seen it fail.”
Another said he wouldn’t ride in a Tesla robotaxi “if you f***ing paid me.”
The former self-driving engineer who spent years reviewing Tesla crash data called the company’s safety claims “bullsh*t”, adding, “Don’t trust Elon on this.”
It comes as Elon Musk continues to make bold promises about where the technology is heading.
Musk recently said at a shareholders meeting Tesla was close to letting drivers take their eyes off the road entirely.
“We almost feel comfortable allowing people to text and drive, which is kind of the killer app,” he told shareholders.
“In the next month or two, we’re going to look closely at the safety statistics, but we will allow you to text and drive, essentially.”
Six months on, Tesla has yet to approve texting while using FSD.
Despite that, Tesla claims FSD is up to 10 times safer than a human driver — a figure Musk and other executives have repeatedly cited.
MORE: Massive change for Aussie favourite
But when Reuters examined how the data were compiled, it found that Tesla’s comparison of its own crash data to federal crash data was flawed.
According to Reuters analysis, Tesla allegedly measures its crash rate using only incidents serious enough to trigger airbag deployments. The automaker then compares that against a federal dataset on crashes which involves a tow truck removing a vehicle.
The EV maker also allegedly benchmarks its vehicles against the average US car, which is significantly older than the average Tesla.
Phil Koopman, an engineering professor at Carnegie Mellon University and autonomous-vehicle safety expert, said the data comes across as skewed because newer vehicles are safer across all brands, regardless of any autonomous features.
“Any new car is dramatically safer than a 12-year-old car,” Phil told the publication.
“It’s like saying: ‘My jet airplane is faster than your World War II bomber.’ Yeah, so, what’s your point?”
MORE: Call to fine Aus drivers of SUVs, utes more
Former Tesla employees also raised questions about how the company presents its technology publicly, particularly around its ride-sharing service, Robotaxi.
They revealed ahead of demonstrations, staff worked long hours mapping routes and training the software on specific hazards, including a Robotaxi pilot in Texas where some human safety monitors rode in the cars while others worked remotely.
If Elon Musk intends to deploy the feature on a large scale, former staffers say these labour-intensive safeguards would be nearly impossible to implement.
Despite this, Musk claims Tesla’s self-driving technology will work globally, relying solely on cameras and AI, and current Tesla owners will be able to get full autonomy through software updates.
In Australia, FSD is available in cars that support HW4 hardware and can handle many driving situations, though it requires driver supervision and has struggled in complex scenarios.
