Robotaxi troubles: what Tesla and Waymo incidents reveal about self-driving

Two of the best-funded robotaxi programs in the world are spending this year fixing problems that were supposed to be solved before paying passengers ever got in. Waymo recalled nearly 4,000 vehicles on June 18 to stop them driving on highways after at least 13 incidents in which its robotaxis entered closed construction zones, according to filings with the National Highway Traffic Safety Administration (NHTSA) reported by TechCrunch. Six of those happened in Phoenix in April and seven in the San Francisco Bay Area in May.

Tesla's robotaxi began limited early access rides in Austin on Sunday, June 22, 2025, for a group of influencers and investors. Business Insider asked three autonomous driving experts to review videos of the rides.

What went wrong on the road

The Tesla issues captured on video, per Business Insider, included speeding more than 10 miles above the limit, driving into the wrong lane, random braking, problems with the pullover button, and at least two interventions by the safety monitor because of parking. Tesla did not respond to a detailed request for comment from Business Insider, and the influencers and investors contacted by the publication did not respond either.

"To process camera data, one has to use AI and machine learning," Raj Rajkumar, a professor of engineering at Carnegie Mellon University, told Business Insider. "But hallucinations are an integral part of how AI operates, and once you hallucinate, phantom braking ends up happening, so a camera-only solution will not be sufficient for a very long time."

Rajkumar's argument is about sensing. Tesla's robotaxi uses an advanced version of its Full Self-Driving software with eight cameras providing 360-degree visual data, and no radar or lidar. Waymo's cars use multiple sensing methods, including lidar, which scans the environment with laser light pulses.

Steven Shladover, lead researcher at the Partners for Advanced Transportation Technology program at the University of California, Berkeley, told Business Insider he is concerned the camera-only approach will eventually lead to passenger injuries without intervention. "Automated driving needs a combination of sensor data from cameras, radars, and lidars, as well as precise localization relative to a high-accuracy digital map of the roadway environment and other data such as the local rules of the road and speed limits," he said.

Phantom braking is not a new subject for Tesla. The company faces a class-action lawsuit over alleged phantom braking in its Autopilot system, which is a driver assistance system distinct from the supervised version of Full Self-Driving. NHTSA opened an investigation into Model 3 and Model Y vehicles in 2022 after more than 750 drivers complained their cars suddenly braked at high speed while on Autopilot; Business Insider reported that investigation and evaluation are still ongoing. Elon Musk said in 2020, on what was then Twitter, that the issue should be fixed in the latest wide release.

Waymo's problem is water and cones

Waymo's troubles are different in kind but similar in pattern: a system that performs well in the common case and stumbles on the edge case, then takes longer than expected to correct it. On May 21, 2026, TechCrunch reported that Waymo had paused service in four cities after robotaxis struggled with heavy rain and flooded roads. One vehicle was spotted driving through a flooded street in Atlanta before getting stuck for about an hour, according to local news reports; it was recovered and removed.

Waymo paused Atlanta and San Antonio while it worked on a solution, then added Dallas and Houston because of severe weather across Texas, a move first reported by Bloomberg News and confirmed by the company to TechCrunch. The company said in a statement that "safety is Waymo's top priority, both for our riders and everyone we share the road with," and that an unoccupied vehicle encountered a flooded road and stopped during intense rain in Atlanta.

The company had already issued a software recall over flooding and admitted it had not finished developing a "final remedy" for avoiding flooded areas. Instead it shipped an update placing "restrictions at times and in locations where there is an elevated risk of encountering a flooded, higher-speed roadway," according to NHTSA documents. It was not enough. Waymo told TechCrunch the Atlanta storm produced rainfall so heavy that flooding began before the National Weather Service had issued a flash flood warning, watch or advisory, and that those alerts are among the signals it relies on.

"NHTSA is aware of this incident, is in communication with Waymo, and will take appropriate action if necessary," a spokesperson for the safety regulator told TechCrunch.

The June recall over highway construction zones followed a similar sequence. Waymo told NHTSA its robotaxis "did not recognize and drove past ramp closure signs into pre-planned freeway construction zones" in Phoenix in mid-April. Its Field Safety Committee restricted freeway operations in the city while a fix was developed. Then on May 18, seven robotaxis drove into highway lanes under active construction in the Bay Area because the software was "prioritizing the avoidance of other freeway hazards and/or failing to recognize the construction zone," the documents show. Waymo suspended all freeway driving the next day and its safety board decided on the recall on June 8. A fix is "currently under development."

Why the fixes keep coming

Waymo says its vehicles have driven more than 170 million miles autonomously and claims a 13x reduction in serious-injury-or-worse crashes compared with human drivers. It is also expanding fast, with plans to launch in more than 20 cities this year including London and Tokyo. That combination, a large safety claim and a rapid rollout, is what turns each edge case into a public test of the technology.

The company has now issued six recalls for its robotaxis. In December it recalled vehicles over illegal behavior around school buses; in May over flooded roads; earlier ones addressed low-speed collisions with chains and gates, telephone poles, and a problem involving towed trucks. The school bus behavior is the subject of active investigations by both NHTSA and the National Transportation Safety Board, and Waymo produced a batch of documents for NHTSA that were redacted to the public. On May 15, NHTSA sent a second document request because the initial response "necessitates that [NHTSA] receive further data and information." A second set of investigations concerns a January 23 incident in which a Waymo robotaxi struck a child in Santa Monica, California; Waymo has said the vehicle braked to around six miles per hour before impact and that the child suffered minor injuries.

Data is the industry's proposed answer, and it is now a business in itself. Uber launched a division called AV Labs in January 2026 to collect driving data for its more than 20 autonomous vehicle partners, including Waymo, Waabi and Lucid Motors, though no contracts were signed at the time. Uber is not returning to building its own robotaxis, which it stopped after one of its test vehicles killed a pedestrian in 2018 and later sold off in 2020. The division started with one Hyundai Ioniq 5 and a team still bolting on lidars, radars and cameras, according to TechCrunch.

Uber's chief technology officer, Praveen Neppalli Naga, framed the effort as broadening access to data rather than monetizing it. "Our goal, primarily, is to democratize this data, right? I mean, the value of this data and having partners' AV tech advancing is far bigger than the money we can make from this," he told TechCrunch. Danny Guo, Uber's VP of engineering, said the lab has to build the data foundation first: "Because if we don't do this, we really don't believe anybody else can."

Partners will not get raw data. Uber plans to process it into what it calls a semantic understanding layer, and to run partners' driving software in "shadow mode" in its cars so that any divergence between the human driver and the software can be flagged back to the partner. Guo said this could help train models to drive more like a human and less like a robot. Uber expects to grow the division to a few hundred people within a year.

None of this settles the underlying argument. Rajkumar's position, that a camera-only system cannot be sufficient for a very long time, is a direct challenge to Tesla's approach. Shladover's is a challenge to any company that treats sensor redundancy as optional. Waymo's recalls are a reminder that even a multi-sensor stack with a decade of operating experience still meets roads, storms and construction crews it cannot always read.

The regulatory picture on Tesla is separate and more serious. An investigation by RTS, published in April 2026, reported that a leak of internal Tesla data showed the company had concealed thousands of serious incidents involving its autonomous driving systems up to 2022, some of them fatal, and that more than 2,400 customer complaints concerned spontaneous acceleration, with the number of accidents exceeding 1,000. Many were marked "unresolved." A jury ordered Tesla to pay more than 243 million dollars in damages to victims, a first in Autopilot-related cases, and a federal judge confirmed the award in late February; Tesla can still appeal. The company has also faced scrutiny from the Department of Justice over whether it misled consumers, and NHTSA has its own investigation. Tesla and Elon Musk did not respond to interview requests from the RTS investigation team.

The common thread is that robotaxi promises are now being audited by the same roads everyone else uses. Waymo's answer is to pause, recall and patch, city by city. Tesla's is to keep the fleet small, with a human safety monitor in the passenger seat, as Musk said before launch. Both are admissions that the last few percent of driving remains the hard part.