Tesla has started offering Robotaxi rides without a safety monitor in Austin, Texas. After yet another set of missed timelines and a full decade of broken promises, Elon Musk is finally getting a version of the “win” he has been desperately seeking. But considering the alarming crash data we have and the evidence of heavy remote monitoring, should we be excited or terrified?
The announcement was made by Musk on X, where he shared a video of someone riding in a Tesla Robotaxi without the usual safety monitor in the front passenger seat:
Tesla VP of Self-Driving Ashok Elluswamy provided some clarity on the rollout strategy, stating:
“Starting with a few unsupervised vehicles mixed in with the broader robotaxi fleet with safety monitors, and the ratio will increase over time.”
This context is important when you consider everything we know about the Tesla Robotaxi program.
Electrek’s Take
Let’s be clear about what we’re witnessing here. This is not Tesla delivering on its original promise. This is Tesla trying to manufacture a win after years of embarrassing missed deadlines.
As I wrote back in June when Tesla launched this service, Tesla’s Robotaxi launch is a dangerous game of smoke and mirrors. Musk badly needed a win with self-driving, and he saw an opportunity to get one in Austin, Texas, a state with some of the most lax regulations regarding autonomous driving.
When the service finally launched, it came with a “safety monitor” in the front seat, which was basically Tesla’s public FSD with the supervising driver being moved to the passenger seat. Musk promised repeatedly throughout 2025, in September, October, November, and December – that the safety monitors would be removed “by the end of the year.”
He technically delivered on that promise, but only for employee testing and internal rides. The paying public? Still getting safety monitors as of now, when the transition is starting as per Elluswamy’s comment.
The crash rate should concern everyone
Here’s what really concerns me. Throughout 2025, we’ve been tracking Tesla’s Robotaxi crashes reported to NHTSA.
Tesla’s Robotaxi fleet has now been involved in at least 8 crashes since June, all while having a safety monitor in the vehicle who should have been preventing additional incidents. An unknown number of additional accidents, as this data is not being shared by Tesla.
Let me put this in perspective: Based on Tesla’s own disclosure that the fleet had traveled 250,000 miles through early November, that’s a crash rate of approximately once every 60,000 miles. The average human driver crashes about once every 500,000 miles.
That means Tesla’s “autonomous” Robotaxis, which are supposed to be the future of safety, are crashing more than 8 times as often as human drivers. And that’s with a trained safety supervisor in the car ready to intervene.
What’s worse, Tesla abuses NHTSA’s capability to redact information in crash reports. Unlike Waymo, which provides substantial detail about incidents, Tesla redacts the “narrative” section of every report, making it impossible to know who was at fault or what actually happened.
Compare this to Waymo, which has over 125 million fully driverless miles and crashes approximately once every 98,600 miles, without any onboard safety monitor. Tesla is still far behind.
The fleet is tiny, and that’s by design
Here’s another thing Tesla doesn’t want you to know: the Robotaxi program in Austin is much smaller than Musk claims.
A 19-year-old engineering student at Texas A&M, Ethan McKanna, reverse-engineered Tesla’s Robotaxi app to track the availability of the network. His tracker found only 32 different Tesla Model Ys, at the time, used in the Robotaxi network, a far cry from Musk’s prediction of “500 vehicles” by year-end.
Even more telling: McKanna’s data shows Tesla is typically running fewer than 10 Robotaxis at the same time, if they’re running any at all.
Remember when Musk said the Robotaxi would “cover half the US population” by the end of 2025? It didn’t. When he said the fleet would grow by more than 10x? It didn’t. When he said there would be over 1,000 Robotaxis in Austin “within a few months” of launching? There weren’t.
Meanwhile, Waymo is clocking 450,000 weekly driverless rides across Austin, Phoenix, San Francisco, Los Angeles, Atlanta, and now Miami. That’s an 80% increase from six months ago, and they plan to expand to 10 more U.S. cities by the end of 2026.
Heavy reliance on remote monitoring
There’s also substantial evidence that Tesla is heavily relying on remote monitoring to make this work.
Even before the Austin launch, I reported that Tesla’s robotaxi fleet would be powered by “plenty of teleoperation” to ensure safety levels because, in Musk’s own words, “we can’t screw up.”
Earlier this year, Tesla released a photo showing its Robotaxi control room with steering wheels visible in the background. Just because there’s no one in the car, it doesn’t mean a human isn’t guiding the vehicle.
This is fine from a safety perspective, in fact, it’s the responsible thing to do. But let’s not pretend this is the same as solving unsupervised self-driving, which Musk has been promising since 2016.
What happens next will be telling
So where does Tesla go from here? Ashok’s statement about “a few unsupervised vehicles mixed in” with a “ratio that will increase over time” tells us a lot about Tesla’s actual confidence level.
This incremental approach is essentially an admission that Tesla isn’t ready to flip the switch entirely. Compare this to Musk’s December claim that “unsupervised is pretty much solved” – if it were truly solved, why the need for a gradual ratio increase?
The key metrics to watch:
- How fast does the ratio actually change? If Tesla is still running mostly supervised vehicles six months from now, that tells you everything about where the technology actually stands.
- What happens to the crash rate? The supervised fleet was already crashing at 10x the rate of human drivers. If unsupervised vehicles start having incidents, we’ll know quickly whether this rollout was premature.
- Does the fleet actually scale? Musk promised 500 vehicles by year-end 2025. They have around 10 cars running consistently.
To Ashok’s credit, the gradual approach is more responsible than what Musk has been promising. It’s similar to how Waymo rolled out, they started with safety drivers, then removed them gradually as they gained confidence. The difference is Waymo did this years ago and now has over 100 million fully driverless miles.
The bottom line
What is the point of mixing in “a few unsupervised vehicles” with your supervised fleet if you already have a crash rate higher than human drivers, with the safety monitor presumably preventing further crashes?
It is concerning to say the least.
Let’s be clear, Tesla is only operating a small fleet to limit the potential for crashes.
The charitable interpretation is that Tesla is being appropriately cautious, gathering data on unsupervised performance in a controlled way before scaling up. That would actually be the responsible approach.
The cynical interpretation, and given Musk’s track record, I lean this way, is that this is about optics. Tesla can now claim to have “unsupervised robotaxis” while the vast majority of vehicles still have safety monitors. It’s a way to manufacture a win after a decade of missed deadlines.
Ashok’s measured statement about gradually increasing the ratio stands in stark contrast to Musk’s claim that “unsupervised is pretty much solved.” If it were solved, you wouldn’t need a gradual ratio increase, you’d just remove all the safety monitors.
The truth is probably somewhere in between. Tesla has made progress. The technology is better than it was. But “better” is not the same as “ready.” Waymo is operating fully driverless commercial services in multiple cities with over 100 million miles of data showing they are safer than humans. Tesla is mixing a few unsupervised vehicles into a fleet of 30-some cars in Austin, most of which still have safety monitors.
That’s not leading the industry. That’s playing catch-up while trying to look like you’re winning.
FTC: We use income earning auto affiliate links. More.
