A Fatal Tesla Crash Shows the Limits of Full Self-Driving

Bloomberg

2 days ago

Industries | The Big Take
As Elon Musk touts robotaxis in Austin, federal regulators are investigating whether the system is dangerous even with a human behind the wheel.
By and Craig Trudell
The setting sun was blinding drivers on the Arizona interstate between Flagstaff and Phoenix in November 2023. Johna Story was traveling with her daughter and a co-worker in a black Toyota 4Runner around a curve that turned directly into the glaring sunlight. They pulled over to help direct traffic around two cars that had crashed.
Back before that curve, Karl Stock was behind the wheel of a red Tesla Model Y. He had engaged what the carmaker calls Full Self-Driving, or FSD — a partial-automation system Elon Musk had acknowledged 18 months earlier was a high-stakes work in progress.
In a few harrowing seconds, the system's shortcomings were laid bare by a tragedy. The Tesla hit Story, a 71-year-old grandmother, at highway speed. She was pronounced dead at the scene.
► Footage from a front camera on the Tesla shows the bright sun setting.
Sun glare becomes more pronounced around the curve.
► A car in the right lane begins to brake. The Tesla appears to maintain its speed.
► More vehicles with hazard lights are parked on the right shoulder.
The Tesla still hasn't braked as they come into view.
► A person waves to get the Tesla driver's attention. The Tesla veers left.
► The Tesla sideswipes the 4Runner and hits Johna Story head-on.
Story's death — one of 40,901 US traffic fatalities that year — was the first known pedestrian fatality linked to Tesla's driving system, prompting an ongoing federal investigation into whether Full Self-Driving poses an unacceptable safety risk. Bloomberg News is publishing photos and partial footage of the crash, which was recorded by the Model Y and downloaded by police, for the first time after obtaining the images and video through a public-records request.
As the National Highway Traffic Safety Administration investigates whether Tesla's human-supervised driving system is potentially defective, the company is proceeding with plans to deploy a small number of vehicles without anyone behind the wheel.
Tesla has already started putting driverless Model Ys on public roads for testing in Austin ahead of plans to launch a driverless taxi service on June 12, Bloomberg reported last week. What regulators still want to know is whether Tesla's cars will be capable of confronting conditions similar to those in the 2023 crash.
A spokesperson for NHTSA said the agency will take any actions necessary to protect road safety. After spending years investigating Autopilot — a different suite of Tesla driver-assistance features — the regulator found that the carmaker hadn't done enough to prevent drivers from misusing the features. Tesla then recalled 2 million cars.
Representatives for Tesla, including Musk, the company's chief executive officer, didn't respond to a list of questions from Bloomberg.
Bryant Walker Smith, a lawyer and engineer who advises cities, states and countries on emerging transportation technologies, warned that Tesla's push to deploy driverless cars may be premature. 'They are claiming they will be imminently able to do something — true automated driving — that all evidence suggests they still can't do safely,' he said.
Johna Story woke up early the morning of Nov. 27, 2023, to take her grandchildren to school. She and her daughter, Sarah, then headed to Phoenix Sky Harbor International Airport, where they worked transporting vehicles for a rental-car company.
That evening, Johna, Sarah and a co-worker, Brian Howard, were driving the Toyota 4Runner from Flagstaff to Phoenix when they came upon an accident shortly after the curve on Interstate 17.
Howard exited the SUV and walked back behind the initial crash scene to alert oncoming traffic. Johna stepped out of the front passenger-side door and put on an orange reflective safety vest.
The Tesla driver, Karl Stock, was traveling at 65 miles per hour in his Model Y, according to the crash report police compiled. As multiple cars ahead began to brake or came to complete stops near the scene of the first crash, the footage appears to show the Tesla maintained its speed.
The Model Y swerved to the left just as it sped past Howard, who stood on the shoulder of the interstate, whipping what appears to be a safety vest in the air. The Tesla jerked back to the right and hit Johna Story head-on with the front bumper, hood and windshield, sending her body tumbling through the air.
'Sorry everything happened so fast,' Stock wrote in a witness statement for police. 'There were cars stopped in front of me and by the time I saw them I had no place to go to avoid them.'
The crash report doesn't mention Full Self-Driving or whether Stock tried to override Tesla's system. Stock wasn't cited, a spokesperson for the Arizona Department of Public Safety said. The victim's family has sued Stock, who didn't respond to requests for comment, and Tesla.
The fatal collision — and the involvement of Full Self-Driving — came to light because of a standing general order from NHTSA that requires Tesla and other carmakers to report crashes involving cars with driver-assistance systems engaged.
Tesla reported the crash seven months after the incident, according to data collected by NHTSA.
Read the full Arizona crash report.
Two months after Story's death in Arizona, a Tesla Model 3 with Full Self-Driving engaged crashed in Nipton, California, in January 2024. Another Model 3 crashed two months later in Red Mills, Virginia, followed by another two months later in Collinsville, Ohio. In all four incidents, the collisions occurred in conditions that reduced roadway visibility, such as sun glare, fog or airborne dust.
The crashes spurred the defect investigation NHTSA opened in October — 18 days before the US election. Musk was campaigning in Pennsylvania that week for Donald Trump and posting regularly about politics on X, his social media network formerly known as Twitter.
A day after NHTSA went public with its investigation, Musk wrote on X that Washington had become 'an ever-increasing ocean of brake pedals stopping progress.'
The following week, Musk took time out of Tesla's quarterly earnings call to advocate for a federal approval process for the deployment of autonomous vehicles to supersede state-by-state regulations.
'If there's a Department of Government Efficiency, I'll try to help make that happen,' he said.
Musk ended up pumping roughly $290 million into the election in support of Trump and the Republicans and became the public face of DOGE, the organization that razed through agencies across Washington.
Trump appointed Sean Duffy to run the Department of Transportation, which oversees NHTSA. Duffy, a former contestant on the MTV reality television series Road Rules, has repeatedly said a federal autonomous vehicle framework is a top priority — both for him and Trump.
On May 20, Duffy visited Musk at Tesla's factory and headquarters in Austin. 'We're here because we're thrilled about the future of autonomous vehicles,' the Transportation Secretary said, standing between Musk and one of Tesla's humanoid robot prototypes. While safety is key, Duffy said in a video posted on X, he also emphasized the need to ' let innovators innovate.'
Last month, NHTSA sent Tesla a letter asking about basic elements of the company's robotaxi plans only weeks before its slated launch.
Tanya Topka, director of NHTSA's Office of Defects Investigation, asked Tesla to state the name of the driving systems the company planned to use in driverless taxis, and to detail the number of vehicles it expected to deploy and in what locations. She asked Tesla to describe each of the sensors the company planned to use for its cars to perceive their surroundings, and how the company intended to ensure safety in conditions where roadway visibility is reduced, including sun glare.
Crashes like the one that claimed Story's life expose the limits of the hardware Tesla relies on for its driving systems, according to Smith, an associate professor at the University of South Carolina. The company uses a third of the number of sensors that Alphabet Inc. 's Waymo vehicles have. Whereas Waymo employs an array of different kinds of sensors — cameras, radar and lidar — Tesla utilizes only one.
Teslas Use Fewer Sensors to See Their Surroundings
Waymo relies on radar and lidar in addition to cameras
'There is a reason why Superman could shoot lasers from his eyes, but also uses his eyes for seeing. You want redundancy,' Smith said. 'It is shocking to many people that Tesla has such confidence in a single type of sensor.'
That single type of sensor Tesla relies on is cameras. The company places an array of them in each vehicle it sells to capture road signs, traffic lights, lane markings and surrounding cars.
Other automakers and autonomous-driving technology companies also use radar and lidar, which emit energy — in lidar's case, laser pulses traveling at the speed of light — to detect the distance of surrounding objects. As active sensors that generate their own signals, they're not affected by external lighting conditions and function better than cameras in direct sunlight.
One advantage cameras have over both radar and lidar is cost. Analysts at BloombergNEF estimated in a report late last year that the sensor suite on a Tesla Model 3 costs just $400. The researcher said that the 24 sensors on the Jaguar I-Pace SUVs that Waymo had deployed in states including Arizona cost 23 times more: roughly $9,300 per vehicle.
'The issue with Waymo's cars is they cost way more money,' Musk said during Tesla's most recent earnings call in April.
Colin Langan, an equity analyst at Wells Fargo, asked Musk later during the call about how Tesla expected to get around the issue of sun glare overwhelming its cameras.
'Actually, it does not blind the camera,' Musk replied, citing a 'breakthrough that we made some time ago.'
Colin Langan: You're still sticking with the vision-only approach. A lot of autonomous people still have a lot of concerns about sun glare, fog, and dust. Any color on how you anticipate on getting around those issues, because my understanding, it kind of blinds the camera when you get glare and stuff. Elon Musk: Actually, it does not blind the camera. We use an approach, which is direct photon count. So, when you see a processed image, so the image that goes from the sort of photon counter -- the silicon photon counter, that then goes through a digital signal processor or image signal processor, that's normally what happens. And then the image that you see looks all washed out, because if you point the camera at the sun, the post-processing of the photon counting washes things out. It actually adds noise. So, part of the breakthrough that we made some time ago was to go with direct photon counting and bypass the image signal processor. And then you can drive pretty much straight at the sun, and you can also see in what appears to be the blackest of nights. And then glare and fog, we can see as well as people can, probably better, but I'd say probably slightly better than people, well, than the average person anyway, and yeah. Langan: So, the camera is able to see when there's direct glare on it; a little surprised by that. Musk: Yeah, yeah.
Musk's comments perplexed Sam Abuelsamid, a former vehicle-development engineer who's now vice president of Telemetry, a Detroit-based communications firm.
'At some point, you have to process the signal,' Abuelsamid said. 'There has to be an image processor somewhere.
Tesla's online owner's manual for the Model Y cautions that the vehicle's front-facing cameras 'may not detect objects or barriers that can potentially cause damage or injury,' and that 'several external factors' can reduce their performance. The company says the cameras aren't intended to replace drivers' visual checks or substitute careful driving.
In other forums, Musk and Tesla have struck a much different tone about the capabilities of its cars.
'Sip tea in a Tesla while it drives you,' the CEO wrote in a May 9 post, sharing a video of an X user filming herself holding a cup on a saucer while in slow-moving traffic.
In another post, the company claimed: 'With FSD Supervised, your steering wheel may start collecting dust.'
The recent posts follow previous warnings from NHTSA about Tesla's social media activity. Shortly after opening its defect investigation, the agency released an email it had sent months earlier, in May 2024, taking issue with Tesla reposting X users who had disengaged from the task of driving while using the company's partial-automation systems.
'While Tesla has the discretion to communicate with the public as it sees fit, we note that these posts show lost opportunities to temper enthusiasm for a new product with cautions on its proper use with the points that Tesla has made to us,' Gregory Magno, a division chief within NHTSA's Office of Defects Investigation, wrote to Tesla.
A NHTSA spokesperson declined to comment specifically on whether the agency had concerns about Tesla's more recent posts, but noted that the company's vehicles aren't self-driving and require a fully attentive driver to be engaged at all times.
Musk has said that 'solving' self-driving is pivotal to Tesla's future. 'That's really the difference between Tesla being worth a lot of money and being worth basically zero,' he said in May 2022.
The CEO has put off plans for a cheaper mass-market vehicle that many Tesla investors — and some insiders — pushed for and view as crucial to the carmaker's near-term prospects.
'Elon Musk has bet the entire company on this philosophy that current Tesla vehicles are capable of being a robotaxi,' said Michael Brooks, executive director of the Center for Auto Safety, a Washington-based advocacy group. 'We know the FSD system is camera-based, and sun glare can inhibit camera-based operations.'
More On Bloomberg
Terms of Service Do Not Sell or Share My Personal Information Trademarks Privacy Policy
Careers Made in NYC Advertise
Ad Choices
Help ©2025 Bloomberg L.P. All Rights Reserved.