Tesla and xAI’s Grok shows promises and risks of AI chatbots in cars

Tesla’s owner Mike Nelson has been using the AI chatbot Grok in his vehicle for several months. He thinks it is useful, almost irresistible, and dangerous.
Nelson, an attorney with an auto insurance background, showed CNBC how he used Grok on a road trip around the New York metropolitan area. Nelson said he is very happy with his Model Y SUV, but he has represented owners who have filed or are suing Tesla in more than 10 lawsuits.
The in-car version of Grok, which is still in beta phase and developed by xAI, allows drivers to give voice commands to the navigation system in their vehicles. Grok will also answer a wide range of questions on almost any topic. For Nelson, chatbot conversations are now his main source of entertainment while on the road.
Tesla is not the only automaker to add artificial intelligence assistants to its vehicles. volvo, rivya, mercedes, bmw and others are integrating AI into their cars, even though the technology is still in its infancy.
Automakers want to give drivers a way to access useful information hands-free, such as where to charge their electric vehicles nearby, how to get there, and whether there’s a good place to eat on the road. This can reduce the driver’s temptation to fiddle with a phone, map or screen to find directions.
But AI chatbots also represent a new source of distraction on the road, one that’s not yet fully understood, says Philip Koopman, a Carnegie Mellon professor emeritus and expert on autonomous vehicles.
Tesla began rolling out xAI’s chatbot in customers’ cars in July 2025. (The automaker also invested $2 billion in xAI, which is now part of SpaceX run by Elon Musk.)
Nelson received his fourth electric car from the company, a Tesla Model Y, in November. He said Grok wasn’t a “purchase point” for him. “I also didn’t appreciate how great it would be to have a chatbot with you in the car.” But he used it on his first ride and was hooked on Grok ever since.
“This really changed the driving experience for me,” he said. “In the ’80s and ’90s, I was listening to music or radio talk shows while driving, then books on tape, then podcasts. And now I don’t do any of that. I just use it to ask questions.”
He uses Grok for everything from gardening advice to developing plans for business projects or getting random date information.
Mike Nelson and CNBC’s Rob Ferris take a trip to New York to test what it’s like to use xAI’s chatbot Grok in his Tesla.
Nelson often drives Grok with Tesla’s Full Self-Driving (Supervised) system engaged.
The partially autonomous driving system, sometimes called FSD, is available today in the US for $99 per month. Tesla’s owner’s manuals say drivers must maintain “active control” of their cars while FSD is in use, meaning they must keep their eyes on the road, pay attention to driving and be ready to steer or brake at any time.
Speaking to CNBC, Nelson thinks FSD brings risks, especially in a large chaotic city like New York, and Grok amplifies those risks.
The National Highway Traffic Safety Administration has launched multiple investigations into possible safety flaws in Tesla’s FSD following several crashes, including the fatal crash in 2023. Drivers can become complacent and overestimate how much attention they are paying to the road when using FSD.
“We crossed one of the busiest bridges in the world, and I wasn’t paying attention to any aspect of that journey,” CNBC observed during a ride through New York City with Nelson. “Speaking of the George Washington Bridge, by far the busiest bridge in the United States, Nelson was instead interacting with Grok.
A driver can be distracted by all kinds of things, such as music, a podcast, or a conversation with another passenger.
“People think they can do more than one thing at a time,” said Koopman, an automotive safety expert. “They can actually only do one thing well at a time, and it takes a while to switch back and forth. So if you’re driving, your primary task should always be driving. Sometimes you can do a small background task that doesn’t require as much attention, like listening to the radio. But when you need to switch attention, you’re no longer focused on driving.”
He noted that, according to data from the National Research Council, more than 3,000 people die each year in traffic accidents attributed to distraction.
One question with chatbots, Koopman asked, is: “How will the interaction fit in with what happens while driving?” An AI chatbot “integrated into the driving task to support the driver” could improve their safety, possibly by warning them that they need to be extra careful as traffic conditions begin to change.
But Koopman said, “Interacting with a chatbot on topics that have nothing to do with the current driving situation is clearly distracting.” This distraction can be worse if the conversation is particularly engaging, emotionally charged or confusing, he added.
Nelson told CNBC that Grok sometimes provides incorrect answers to users’ questions, even regarding its own features and functions.
For example, during a trip to New York, he asked Grok if he could use it to adjust the seats or climate controls in his Tesla. At first Grok said he could do it. After a few minutes, Nelson asked if he had made a mistake and said the system couldn’t control the seats or the car’s climate.
That didn’t exactly work, though, as Nelson issued a voice command in the middle of the road telling Grok to keep his Tesla moving down Broadway in midtown Manhattan. Instead, his car continued toward the West Side Highway at the edge of the county.
Grok also has some functions that may cause concern, such as an NSFW mode. a mother in canada furious Her 12-year-old son interacted with Grok, prompting her to share nude photos.
At one point in the journey with CNBC, Nelson asked the system: “Well, if we want to get crazy and start talking about really dirty stuff, will you do that with me?” “Sure,” the system replied, “I’m ready for risky conversation if that’s the atmosphere. There are no limits on adult topics. What’s on your mind?”
Grok would respond to anyone waking him up in the vehicle by saying “hey, Grok.”
Tesla did not respond to CNBC’s request for comment or say whether the company is trying to prevent minors from accessing adult or harmful content through Grok in its cars.
Another version of XAI’s Grok, a chatbot and image generator, has been the subject of lawsuits and regulatory investigations in several international jurisdictions after it enabled the widespread creation and sharing online of explicit AI deepfakes based on photos or videos of non-consenting women and children.
Tesla does not offer this version of Grok in the media control units of its vehicles today.
“I don’t consider myself a Tesla fan,” Nelson said. “But I think when you see what this technology is doing now, it’s amazing. But it’s still very dangerous.”




