Tesla robotaxis crash within days of Austin pilot launch

[Daniel Sims]

Recap: Controversy has long surrounded Tesla's self-driving technology, which has faced accusations of false advertising, deadly accidents, and allegations of cover-ups. An early report from the company's robotaxi service in Austin, Texas, is unlikely to improve matters, as it reveals that the vehicles were involved in accidents almost immediately after they began operating in late June.

A report to the federal government reveals that Tesla's robotaxi fleet in Austin suffered three crashes soon after the service began on June 23. Forbes reports that the company's data is vague and heavily redacted, but one or more of the accidents might have occurred on the first day.

A Tesla earnings call suggests that all three July incidents happened within the fleet's first 7,000 miles of operation, indicating an extremely poor driving record. For comparison, although rival Waymo reported many more crashes during the period, the company's record after 96 million miles on the road surpasses most human drivers.

Here's my entire first Tesla Robotaxi trip from start to finish. Ten miles across south Austin. Timestamps in comments. pic.twitter.com/cr6y4z1UWB

– Rob Maurer (@TeslaPodcast) June 22, 2025

Two vague Tesla reports describe rear impacts, one occurring during a right turn and the other after the robotaxi stopped while driving straight for unspecified reasons. It remains unclear whether the Teslas were at fault in the incidents. However, another crash involved a robotaxi that struck a stationary object, causing minor injuries and requiring the Tesla to be towed.

Videos showing the robotaxis behaving erratically appeared on the first day of the Austin trial, prompting an investigation by the National Highway Traffic Safety Administration. Videos show the Teslas speeding and breaking traffic laws.

Notably, the clips came from Tesla investors. The Austin pilot program is currently limited to influencers, the company's fans, and any adult guests. Furthermore, human supervisors accompany the passengers and can activate emergency brakes from passenger seats, making the accidents more concerning.

Here is my experience in one of the first public nighttime @Tesla Robotaxi rides. It was smooth, comfortable and just as good as it is during the daytime. pic.twitter.com/7bDvVaWf3A

– Sawyer Merritt (@SawyerMerritt) June 23, 2025

Multiple companies have found the robotaxi business challenging. Waymo's early trials prompted an NHTSA investigation following numerous incidents, and General Motors quit robotaxis after encountering similar challenges.

Moreover, electric vehicle drivers are not enthusiastic about full automation. An August survey indicated that most Tesla customers were either indifferent or hostile to the company's Full Self-Driving technology. Nearly half of the respondents believe it should be illegal.

Despite the setbacks, robotaxis continue to spread. Wayve and Uber intend to begin operating in the United Kingdom sometime next year, and Nvidia is reportedly planning to step into the arena.

Permalink to story:

Tesla robotaxis crash within days of Austin pilot launch

 

[Megalomaniac]

There are literally people arguing with me when I say that "Robotaxi will NEVER work in New York City".

How do they ignore these stories?

I've used FSD in the Cyberbeast. While it's impressive, it doesn't understand the etiquette of driving. It doesn't understand that when you're 17 feet long, you can block traffic moving behind you - it's only concerned with what's ahead. It doesn't understand when to slow down for threats or obstacles it can't perceive. My GM Supercruise is better and won't work in situations that are dangerous. FSD is pushing the limits of safety too far.

 

[wiyosaya]

I'm shocked! Totally shocked. Tesla, with multiple known and widely-publicized accidents with people using "Autopilot" including crashing into parked emergency vehicles, has its robo-taxis crash as soon as the vehicles go into service and some of the crashes are into stationary objects? No, it isn't so!!!!!! /s

 

[VitalyT]

Too many things on the road do not make any logical sense, including behavior of pedestrians and other drivers. It would take an AI that's smarter than human to provide a safe service. Until then, it will always be a disaster. We are at least 10 years away before AI taxis can be widely accepted.

 

[wiyosaya]

There are literally people arguing with me when I say that "Robotaxi will NEVER work in New York City".

How do they ignore these stories?

Yes. They should be laughed off the face of the earth!!

 

[wiyosaya]

Too many things on the road do not make any logical sense, including behavior of pedestrians and other drivers. It would take an AI that's smarter than human to provide a safe service. Until then, it will always be a disaster. We are at least 10 years away before AI taxis can be widely accepted.

Yes, like we are 20-years away from Fusion!

 

[Plutoisaplanet]

Moreover, electric vehicle drivers are not enthusiastic about full automation. An August survey indicated that most Tesla customers were either indifferent or hostile to the company's Full Self-Driving technology. Nearly half of the respondents believe it should be illegal.

This is highly deceptive. I recommend not citing this. Here's why:

They were not Tesla customers as you say. They were "consumers" in general "weighted by education, race, gender, age, income, geography, and political preference". We were not told what the weights were (the report didn't even say they were weighted to match EV buyers).

They were also not shown Tesla's marketing. Tesla's marketing names the feature "Full Self-Driving (Supervised)" while the survey says that the feature is named "Full Self-Driving". So even worse, they're misrepresenting Tesla's marketing, using a different name that probably would be illegal. See for yourself that in this page alone, Tesla names the feature over 10 times as "Full Self-Driving (Supervised)": https://www.tesla.com/fsd

 

[Impudicus]

Too many things on the road do not make any logical sense, including behavior of pedestrians and other drivers. It would take an AI that's smarter than human to provide a safe service. Until then, it will always be a disaster. We are at least 10 years away before AI taxis can be widely accepted.

I've seen how some humans drive.

 

[ScottSoapbox]

Tesla has an uphill battle with the decision to focus more on cameras than LIDAR. If they can crack it, there is a manufacturing cost savings, but they are still working on the computer model being smart enough.

 

[AnilD]

The fact that nearly half of Tesla owners think Full Self Driving should be illegal while the company is simultaneously launching a robotaxi fleet says a lot about the disconnect between marketing and reality.

Three crashes in the first 7,000 miles is basically like buying a new Roomba and it immediately runs full speed into your furniture on day one but this time the furniture is a person’s car.

 

[wiyosaya]

Tesla has an uphill battle with the decision to focus more on cameras than LIDAR. If they can crack it, there is a manufacturing cost savings, but they are still working on the computer model being smart enough.

Personally, I think the computer not being smart enough reflects that some people in Tesla are not smart enough.

 

[wiyosaya]

This is highly deceptive. I recommend not citing this. Here's why:

They were not Tesla customers as you say. They were "consumers" in general "weighted by education, race, gender, age, income, geography, and political preference". We were not told what the weights were (the report didn't even say they were weighted to match EV buyers).

They were also not shown Tesla's marketing. Tesla's marketing names the feature "Full Self-Driving (Supervised)" while the survey says that the feature is named "Full Self-Driving". So even worse, they're misrepresenting Tesla's marketing, using a different name that probably would be illegal. See for yourself that in this page alone, Tesla names the feature over 10 times as "Full Self-Driving (Supervised)": https://www.tesla.com/fsd

Personally, I think the nits that you are picking would not make much difference to those who were surveyed.

IMO, what you are saying is along the lines of Steve Jobs, in response to that one iPhone that had antenna problems, saying "You're holding it wrong."

Imagine: Tesla rep: "OMG, you crashed???? You're driving it wrong!!!!"

 

[ScottSoapbox]

Personally, I think the computer not being smart enough reflects that some people in Tesla are not smart enough.

To be fair, the people at Tesla are attempting to solve a more difficult problem than those using LIDAR.

But that strategy's tradeoffs (harder problem for cheaper hardware) probably seemed better when they started and there was much less competition in the space.

 

[Plutoisaplanet]

Personally, I think the nits that you are picking would not make much difference to those who were surveyed.

IMO, what you are saying is along the lines of Steve Jobs, in response to that one iPhone that had antenna problems, saying "You're holding it wrong."

Imagine: Tesla rep: "OMG, you crashed???? You're driving it wrong!!!!"

Wiyosaya 2025: It's not a driver's fault for crashing, it's Tesla's.

 

[jeffbert]

Johnny Cab had no such problems; yet, it does require the passenger specify a definite destination. Saying, "Just go!" gets you nowhere.

 

[DrSuess]

Too many things on the road do not make any logical sense, including behavior of pedestrians and other drivers. It would take an AI that's smarter than human to provide a safe service. Until then, it will always be a disaster. We are at least 10 years away before AI taxis can be widely accepted.

Yeah, but Waymo has been operating in my area under those same conditions and is not having issues, even on narrow or congested streets.

 

[JSabin]

More than one long-time automobile industry journalist has proclaimed self-driving cars will never be reliable until government installs wires in roadways delineating lanes. Video cameras and Lidar will never cut it. Think they are right.

 

[wiyosaya]

To be fair, the people at Tesla are attempting to solve a more difficult problem than those using LIDAR.

But that strategy's tradeoffs (harder problem for cheaper hardware) probably seemed better when they started and there was much less competition in the space.

Yes, like I said Some people in Tesla are not smart enough.

 

[wiyosaya]

And @ScottSoapbox @Plutoisaplanet

Yeah, but Waymo has been operating in my area under those same conditions and is not having issues, even on narrow or congested streets.

 

[captaincranky]

They were also not shown Tesla's marketing. Tesla's marketing names the feature "Full Self-Driving (Supervised)" while the survey says that the feature is named "Full Self-Driving". So even worse, they're misrepresenting Tesla's marketing, using a different name that probably would be illegal. See for yourself that in this page alone, Tesla names the feature over 10 times as "Full Self-Driving (Supervised)"

So basically what you're saying is, "these robotaxis should have never been sent out without a human "supervisor" onboard, in the first place".
Which lays all the blame of,` "misrepresentation" you're babbling about, squarely on Elon Musk and Tesla's shoulders.

Although, with who they have for a governor, senator, and who they voted far as president, Texans do present a bit on the stupid side.

Hey, maybe they'll be able to "gerrymander" a district where they can run these clunkers without them crashing.

 

[Plutoisaplanet]

So basically what you're saying is, "these robotaxis should have never been sent out without a human "supervisor" onboard, in the first place".
Which lays all the blame of,` "misrepresentation" you're babbling about, squarely on Elon Musk and Tesla's shoulders.

Although, with who they have for a governor, senator, and who they voted far as president, Texans do present a bit on the stupid side.

Hey, maybe they'll be able to "gerrymander" a district where they can run these clunkers without them crashing.

I guess you aren’t aware but all robotaxis DO have a supervisor on board currently.

Also, robotaxis have separate software from FSD Supervised:

https://twitter.com/x/status/1947055949992321437

 

[captaincranky]

I guess you aren’t aware but all robotaxis DO have a supervisor on board currently.

Also, robotaxis have separate software from FSD Supervised:

https://twitter.com/x/status/1947055949992321437

So what exactly do the "supervisors" do? Get out, take pictures, and estimate the damage after the crash?

Those two "tweets" don't do much beyond proving that Musk is still, as always, full of sh!t
Seriously, if Austin already got the, "new and improved, good software", Europe and China better brace themselves.

 

[Plutoisaplanet]

So what exactly do the "supervisors" do? Get out, take pictures, and estimate the damage after the crash?

Those two "tweets" don't do much beyond proving that Musk is still, as always, full of sh!t
Seriously, if Austin already got the, "new and improved, good software", Europe and China better brace themselves.

Tesla now puts their robotaxi safety monitors in the driver's seat

Your self-driving car has arrived…but there might be a human at the wheel.

mashable.com

Basically, human safety monitors are required to be present inside a Tesla robotaxi to supervise the driverless vehicle, make sure the software is performing as expected, and intervene if necessary.

 

[captaincranky]

@Plutoisaplanet OK, first, taxi drivers do struggle for their pay. A worst case scenario is an immigrant Indian MD, who has to drive a cab to feed his family, until he (or she) manages to get licensed in the US.

Do remember all the sh!t blown up by taxi drivers over money, when Uber started to take hold?
So now, you have "robo-taxis", which are very likely quite a bit more expensive that human driven vehicles of the same type and capacity,. But you need "driver supervisors" added to the mix, since the cars can't quite be trusted, at least of yet.

So, how much are we going to compensate these supervisors? If' it's on a typical silicon valley pay scale, a ride is going to cost the big bucks. I imagine musk can convince some people to compensate him on the premise that this is not merely a taxi ride, but a fun filled amusement park experience. Sic: "come see the magic taxi drive itself". However, many will still opt for good old public transit. In other words, they'll, "take the bus".

 

[Plutoisaplanet]

@Plutoisaplanet OK, first, taxi drivers do struggle for their pay. A worst case scenario is an immigrant Indian MD, who has to drive a cab to feed his family, until he (or she) manages to get licensed in the US.

Do remember all the sh!t blown up by taxi drivers over money, when Uber started to take hold?
So now, you have "robo-taxis", which are very likely quite a bit more expensive that human driven vehicles of the same type and capacity,. But you need "driver supervisors" added to the mix, since the cars can't quite be trusted, at least of yet.

So, how much are we going to compensate these supervisors? If' it's on a typical silicon valley pay scale, a ride is going to cost the big bucks. I imagine musk can convince some people to compensate him on the premise that this is not merely a taxi ride, but a fun filled amusement park experience. Sic: "come see the magic taxi drive itself". However, many will still opt for good old public transit. In other words, they'll, "take the bus".

Obviously the plan for any autonomous taxi service is to make it cheaper than everything besides public transit while offering privacy. The safety monitors will be removed once Robotaxis are safe enough to constantly operate on their own. If you’re so curious about them, why don’t you learn more about them?

And fyi Tesla plans to start mass production of a $25k cybercab next year.