- Private camera footage captured by cars shared in chat rooms: ex-workers
- Clips circulated included one of a child being hit by a car: ex-employee
- Tesla Says Vehicle Camera Footage ‘Remains Anonymous’
- A video showed a submersible from the James Bond film, owned by Elon Musk
LONDON/SAN FRANCISCO, April 6 (Portal) – Tesla Inc assures its millions of electric car owners that their privacy “is and always will be hugely important to us”. The cameras it puts in vehicles to aid driving are “designed from the ground up to protect your privacy,” according to its website.
But between 2019 and 2022, groups of Tesla employees privately shared sometimes highly invasive videos and images captured by customers’ dash cameras via an internal messaging system, according to Portal interviews with nine former employees.
Some of the footage caught Tesla customers in embarrassing situations. A former employee described video of a man approaching a vehicle completely naked.
Also shared: Accidents and incidents on the road. A crash video in 2021 showed a Tesla traveling at high speed in a residential area and hit a child on a bicycle, according to another ex-employee. The kid flew one way, the bike the other. The video spread “like wildfire” over private one-on-one calls at a Tesla office in San Mateo, California, the ex-employee said.
Other images were more mundane, like images of dogs and funny street signs, which employees memes by dressing them up with amusing captions or comments before posting them in private group chats. While some posts were only shared between two employees, others could be seen by dozens of them, according to several former employees.
Tesla states in its online “Customer Privacy Policy” that its “camera footage remains anonymous and is not linked to you or your vehicle.” But seven former employees told Portal the computer program they used at work could show the location of the records – potentially revealing where a Tesla owner lived.
A former employee also said some footage appears to have been taken as cars were parked and turned off. A few years ago, Tesla received video footage of its vehicles, even when they were off, if the owners gave their consent. Since then it has stopped.
“We could see into people’s garages and people’s homes,” said another former employee. “Let’s say a Tesla customer had something special in their garage, you know, people would post things like that.”
Tesla has not responded to detailed questions sent to the company for this report.
About three years ago, some employees came across video of a one-of-a-kind submersible parked in a garage and shared it, according to two people who viewed it. Nicknamed Wet Nellie, the white Lotus Esprit Sub starred in the 1977 James Bond film The Spy Who Loved Me.
The vehicle was owned by Tesla CEO Elon Musk, who bought it at auction in 2013 for around $968,000. It’s not clear if Musk knew of the video or shared it.
Musk did not respond to a request for comment.
To cover this story, Portal contacted more than 300 former Tesla employees who had worked for the company over the past nine years and were involved in the development of its self-driving system. More than a dozen agreed to answer questions, all speaking on condition of anonymity.
Portal was unable to obtain any of the released videos or images, which former employees said they had not kept. The news agency also couldn’t determine whether the record-sharing practice, which existed in some parts of Tesla as recently as last year, continues today or how widespread it was. Some former employees who were contacted stated that the only disclosure they observed was for legitimate work purposes, e.g. B. to seek help from colleagues or superiors.
IDENTIFICATION OF PEDESTRIANS AND ROAD SIGNS
Sharing sensitive videos illustrates one of the less-noticed characteristics of artificial intelligence systems: they often require armies of humans to help train machines to learn automated tasks like driving.
Since around 2016, Tesla has employed hundreds of people in Africa and later the United States to annotate images so its cars can learn to recognize pedestrians, street signs, construction vehicles, garage doors, and other objects encountered on the road or at customer sites . To achieve this, data labelers were given access to thousands of videos or images captured by car cameras to view and identify objects.
Tesla has increasingly automated the process, closing a data labeling center in San Mateo, California, last year. But it continues to employ hundreds of record labelers in Buffalo, New York. In February, Tesla announced that its workforce had grown 54% to 675 over the previous six months.
Two former employees said they didn’t mind sharing pictures, that customers had given their consent, or that people long ago gave up any reasonable expectation of keeping personal information private. However, three others said they were concerned.
“It was an invasion of privacy, to be honest. And I always joked that I would never buy a Tesla after seeing how they treat some of these people,” said a former employee.
Another said: “It bothers me because the people buying the car don’t know their privacy isn’t being respected… We could see them doing laundry and really intimate things. We could see their children.”
A former employee saw nothing wrong with sharing images, but described a feature that allowed data identifiers to show the location of recordings on Google Maps as a “massive invasion of privacy”.
David Choffnes, executive director of the Cybersecurity and Privacy Institute at Northeastern University in Boston, described Tesla employees’ sharing of sensitive videos and images as “morally reprehensible”.
“Any normal person would be appalled,” he said. He pointed out that the distribution of sensitive and personal content could be construed as a violation of Tesla’s own privacy policy – potentially leading to intervention by the US Federal Trade Commission, which enforces federal laws related to consumer privacy.
A spokesman for the FTC said it does not comment on individual companies or their conduct.
To develop the technology for self-driving cars, Tesla collects a vast amount of data from its global fleet of several million vehicles. The company requires car owners to provide permission on their cars’ touchscreens before Tesla collects their vehicles’ data. “Your data is yours,” says Tesla’s website.
In its customer privacy policy, Tesla states that if a customer consents to the sharing of data, “Your vehicle may collect the data and provide it to Tesla for analysis. This analysis helps Tesla improve its products and features and diagnose problems more quickly.” It also said the data may include “short video clips or images,” but is not linked to a customer account or vehicle identification number, “and not you personally identify”.
Carlo Piltz, a data protection lawyer in Germany, told Portal it is difficult to find legal justification under European data protection law for sharing vehicle records internally when it has “nothing to do with providing a safe car or the functionality” of the self-driving system by Tesla.
In recent years, Tesla’s dashcam system has sparked controversy. In China, some government buildings and neighborhoods have banned Teslas over camera concerns. In response, Musk said in a virtual talk at a Chinese forum in 2021, “If Tesla uses cars to spy in China or anywhere else, we will be shut down.”
Elsewhere, regulators have been investigating the Tesla system for potential data breaches. But the privacy cases tended not to focus on the rights of Tesla owners, but rather on bystanders who were unaware they could be recorded by parked Tesla vehicles.
In February, the Dutch Data Protection Authority (DPA) announced that it had closed an investigation into Tesla over possible data breaches related to “Sentry Mode,” a feature designed to record suspicious activity while a car is parked and alert the owner to warn.
“People walking past these vehicles were filmed without knowing it. And the owners of the Teslas could go back and look at these pictures,” said dpa board member Katja Mur in a statement. “If a person parked one of these vehicles in front of someone’s window, they could spy in and see everything the other person was doing. This is a serious invasion of privacy.”
The watchdog determined that it wasn’t Tesla, but the owners of the vehicles who were legally responsible for their car’s records. It said it decided not to penalize the company after Tesla said it made several changes to Sentry mode, including pulsing a vehicle’s headlights to alert passers-by they might be being recorded.
A DPA spokesman declined to comment on Portal’ findings, but said in an email: “Personal data must be used for a specific purpose and sensitive personal data must be protected.”
REPLACING HUMAN DRIVERS
Tesla calls its automated driving system Autopilot. Launched in 2015, the system included such advanced features as changing lanes with the tap of a turn signal and parallel parking on command. To make the system work, Tesla first installed sonar sensors, radar, and a single front-facing camera at the top of the windshield. A follow-up version, introduced in 2016, included eight cameras around the car to collect more data and offer more features.
Ultimately, Musk’s future vision is to offer a “Full Self-Driving” mode that would replace a human driver. Tesla began rolling out an experimental version of this mode in October 2020. Despite requiring the driver to keep their hands on the wheel, it currently offers features such as the ability to automatically slow a car as it approaches stop signs or traffic lights.
Tesla’s autopilot system
This excerpt from the Tesla Model X owner’s manual explains the car’s autopilot system, including the cameras that record video of the vehicle’s surroundings. Portal found Tesla employees shared clips capturing sensitive and embarrassing personal moments.
In February, Tesla recalled more than 362,000 U.S. vehicles to update its Full Self-Driving software after the National Highway Traffic Safety Administration said it could allow vehicles to exceed speed limits and potentially cause intersection accidents.
As with many artificial intelligence projects, Tesla hired data labelers to develop Autopilot to identify objects in images and videos and teach the system how to respond when the vehicle is on the road or parked.
Tesla initially outsourced data labeling to a San Francisco-based nonprofit then known as Samasource, people familiar with the matter told Portal. The organization had an office in Nairobi, Kenya and specialized in providing training and employment opportunities to disadvantaged women and youth.
In 2016, Samasource provided about 400 workers for Tesla there, up from about 20 originally, according to a person familiar with the matter.
However, by 2019, Tesla was no longer satisfied with the work of Samasource’s data writers. At an event called Tesla AI Day in 2021, Andrej Karpathy, then Senior Director of AI at Tesla, said: “Unfortunately, we realized very quickly that working with a third party to get datasets for something so critical just wouldn’t cut it it… Honestly, the quality wasn’t amazing.”
A former Tesla employee said of the Samasource taggers, “They would highlight fire hydrants as pedestrians… They would miss objects all the time. Her ability to draw boxes was very poor.”
Samasource, now called Sama, declined to comment on his work for Tesla.
Tesla decided to introduce data labeling internally. “Over time, we have grown into a data labeling organization (organization) of more than 1,000 people, full of professional labelers who work very closely with the engineers,” Karpathy said in his August 2021 presentation.
Karpathy did not respond to requests for comment.
Tesla’s own data labelers initially worked in the San Francisco Bay Area, including the San Mateo office. Groups of data markers have been assigned a variety of different jobs, including marking road lane lines or emergency vehicles, former employees said.
At one point, Teslas on Autopilot had trouble reversing out of garages and became confused when hitting shadows or objects like garden hoses. Therefore, some data labelers have been asked to identify objects in videos recorded in garages. The problem was eventually resolved.
In interviews, two former employees said that in the course of their normal work duties, they were sometimes asked to look at pictures of customers in and around their homes, including garages.
“I’ve sometimes wondered if these people know we’re seeing this,” one said.
“I’ve seen some scandalous stuff at times, you know, like I’ve seen scenes of intimacy but no nudity,” said another. “And there was definitely a lot of things that seemed like I didn’t want anyone to know about my life.”
As an example, this person recalled seeing “embarrassing items,” like “certain lingerie items, certain sexual wellness items… and just private scenes of life that we were really privy to because the car was charging.”
Memes in the San Mateo office
Tesla staffed his San Mateo office with mostly young workers in their 20s and early 30s, who brought with them a culture that valued fun memes and viral online content. Former employees described a casual atmosphere in chat rooms where workers exchanged jokes over pictures they viewed while tagging.
According to several ex-employees, some labelers have shared screenshots, sometimes tagged with Adobe Photoshop, in private group chats on Mattermost, Tesla’s internal messaging system. There they would attract responses from other workers and managers. Participants would also add their own tagged images, jokes, or emojis to keep the conversation going. Some of the emojis were created specifically to indicate inside jokes from the office, several former employees said.
A former labeler described sharing images as a way to “break the monotony.” Another described how sharing won the admiration of peers.
“If you saw something cool that would get a reaction, post it, right, and later, during the break, people would come up to you and be like, ‘Oh, I saw what you posted. That was funny,'” said this former labeler. “People who were promoted to leadership positions shared a lot of these funny things and gained a reputation for being funny as a result.”
Some of the content shared resembled memes on the internet. There were dogs, interesting cars, and clips of people tripping and falling caught by Tesla cameras. There was also disturbing content, such as someone seemingly being dragged into a car against their will, an ex-employee said.
Video clips of Tesla accidents were also sometimes shared in private chats on Mattermost, several former employees said. These included examples of bad driving or collisions with people who were hit while riding a bicycle – like the one with the child – or motorcycling. Some data identifiers rewind such clips and play them in slow motion.
At times, Tesla executives have cracked down on inappropriate sharing of images on public Mattermost channels, claiming the practice violates company policy. Still, screenshots and memes based on them continued to circulate through private chats on the platform, several ex-employees said. Workers only shared them individually or in small groups in the middle of last year.
One of the perks of working for Tesla as a data writer in San Mateo was the chance to win a prize — use of a company car for a day or two, according to two former employees.
But some of the lucky winners got paranoid about driving the electric cars.
“Knowing how much data these vehicles can collect definitely made people nervous,” said a former employee.
Reported by Steve Stecklow and Waylon Cunningham in London and Hyunjoo Jin in San Francisco. Edited by Peter Hirschberg.
Our standards: The Trust Principles.
Steve Stecklow