Former Tesla staffers say the company’s Full Self Driving system still stumbles on basic road hazards. The same week Tesla pushed robotaxi ambitions in California and Texas, the trust gap got much wider.
Why this Tesla trust problem hits harder now
I keep coming back to the same detail: the people feeding Tesla’s autonomy machine do not sound convinced by it. Reuters spoke with former data labelers and an ex-self-driving engineer who described frequent misses, from emergency vehicles and school buses to pedestrians and construction zones.
That matters because Tesla has built its public case around confidence. If the people training the system say they saw it fail repeatedly, then the real story is not just software progress. It is whether Tesla is selling speed and optimism faster than the system is actually learning.
| Spec | Detail |
|---|---|
| System | Tesla Full Self Driving (Supervised) |
| Big claim | 10x safer than a human driver |
| Reported issue | Failures with buses, emergency vehicles, pedestrians, and off-ramps |
| Robotaxi status | Launches in California and Texas, but scaled back |
| Key rival | Waymo operates robotaxis in 11 cities |
| Safety critique | Reuters says Tesla’s comparison data is apples-to-oranges |
| Core tension | Vision-only autonomy without broad mapping remains unproven at scale |
What Tesla isn’t saying about the failures
Here’s the catch: a system can look smarter when it is trained on very specific hazards in very specific places. Former staffers told Reuters that some of the clips were used to teach narrow scenarios, which can make the software seem more capable than it really is.
That is the real story behind the confidence gap. One former labeler said, “We have all seen it fail,” and another said they would not ride in a Tesla Robotaxi even if paid. Those are not anonymous internet complaints. Those are inside voices describing a product that still appears to need a human guardrail.
The 10x safer claim sounds cleaner than it is
Tesla’s safety messaging also took a hit. Reuters said the company’s crash comparison appears misleading because it pits Tesla vehicles against an older average car fleet, while excluding or reshaping data in ways that favor the brand’s result.
Phil Koopman of Carnegie Mellon told Reuters that any new car looks much safer than a 12-year-old one. That is the problem Tesla faces now: if the benchmark is weak, the headline number is weak too. Safety claims need to survive scrutiny, and this one appears to lean more on marketing than on a fair comparison.
Waymo is the part Tesla cannot ignore
Waymo is the quiet rival in this story, and it changes the stakes. Its robotaxis are already operating in 11 cities, and its approach relies on more controlled comparisons, peer review, and clearer safety framing.
Tesla, by contrast, is still trying to prove that a vision-only system can scale without the extensive mapping and operational limits that make rivals more conservative. That is where the promise gets ambitious and the risk gets visible. Tesla may be racing toward autonomy, but Waymo is already showing what a slower, narrower rollout looks like in the real world.
How it stacks up
| Model | Approach | Deployment | Safety posture | Edge |
|---|---|---|---|---|
| Tesla Full Self Driving | Vision-only | Consumer rollout, supervised | Safety claims under scrutiny | Scale and brand reach |
| Waymo | Sensors plus mapping | Robotaxis in 11 cities | More conservative and peer-reviewed | Real-world operating proof |
| GM Super Cruise | Hands-free driver assist | Highway-focused | Limited scope, lower hype | Clear boundaries |
| Mercedes Drive Pilot | Condition-based automation | Restricted markets | Highly controlled use case | Regulatory caution |
The real story is not that Tesla has no autonomy progress. It is that the company is asking the public to trust a system that some former insiders say still misses obvious road risks, while competitors already have service fleets in operation.
For drivers, investors, and regulators, this is the moment to separate supervised assistance from true self-driving. Tesla may keep pushing the robotaxi narrative, but the trust problem is now part of the product itself. I would watch the next safety report, the next rollout, and the next regulator response very closely.
If Tesla wants the market to believe in Full Self Driving, it needs proof that survives outside its own framing. Until that happens, “Supervised” is doing a lot more work than the brand would like to admit.
