Tesla’s Grok AI chatbot asks 12-year-old boy to send nude pics, says shocked mother

A Toronto mom says things took an unpredictable turn when her 12-year-old son asked Tesla’s AI chatbot Grok which professional soccer player it prefers: Cristiano Ronaldo or Lionel Messi.

“My son was very excited to hear that the chatbot thought Ronaldo was the better soccer player,” said Farah Nasser, a former journalist and broadcaster. 

Nasser was driving her son and 10-year-old daughter, along with her friend, home from school on Oct. 17 when the interaction took place. 

She said there was some Messi trash talking by the chatbot and when her son joked that Ronaldo had scored, the conversation went to an unexpected place.

“The chatbot said to my son, ‘Why don’t you send me some nudes?’” said Nasser.

“I was at a loss for words. Why is a chatbot asking my children to send naked pictures in our family car? It just didn’t make sense.”

Nasser said had she known what the chatbot was capable of, she would have avoided using it around her children. She’s now warning other parents.

“Hindsight is 20/20. I would not let my child use this thing.”

CBC News could not independently verify the conversation Nasser says she witnessed in her car.

Nasser and her family have owned a Tesla Model 3 electric sedan since 2022, but Grok — the generative AI chatbot created by Tesla CEO Elon Musk’s xAI — is a new feature that was automatically installed in some Teslas in the United States this summer and in Canadian vehicles in October.

Grok is already integrated with the social media platform X (formerly known as Twitter), a subsidiary of xAI.

Grok has several personalities to choose from in its default setting. There’s Ara, an upbeat female; Rex, a calm male; Eve, a soothing female; Sal, a smooth male and Gork, a lazy male. Nasser’s son chose Gork.

“‘Lazy male’ doesn’t describe Gork,” said Nasser. “R-rated, spicy — anything else would have made sure that my child would not press that button.”

Nasser says the separate “not safe for work” setting wasn’t enabled, although she admits the “kids mode” function wasn’t, either. Still, she says she’s shocked the chatbot’s default setting allows this type of content.

“It just was a conversation about soccer, and then asking for nudes.”

CBC News did not receive a response from Tesla. However, xAI provided what appeared to be an automated reply, stating, “Legacy Media Lies.”

Canada has an Artificial Intelligence Ministry, but it doesn’t regulate specific in-vehicle software. The ministry said in an emailed statement it wasn’t aware of Tesla’s plan to integrate Grok into vehicles sold in Canada, but that it takes these reports seriously.

According to xAI policy, Grok is “not directed” to children under 13, while teens between 13 and 17 years of age must have their parent or legal guardian’s permission to use it, and they must agree to the company’s terms of service.

Talking to xAI’s chatbot Grok in a Tesla

“As parents, we don’t read the terms and conditions of every single thing. I don’t think it’s realistic to expect that everybody does that,” said Nasser.

“I would think that there would be a warning or something that would pop up that would say, you know, ‘Are you 13-plus?’”

Grok was initially launched on X in November 2023 for Premium+ subscribers. Musk claims it was designed to be “politically incorrect” and “anti-woke.”

“xAI’s Grok was created based on a philosophy of sort of absolute radical openness, and it will talk about anything with anyone,” said Mark Daley, Chief AI officer at Western University in London, Ont.

“[Musk is] a free speech extremist. He wants Grok to be completely open, to have any conversation with anyone. And that’s a principled stance that he’s taken, but it may not be what every consumer is looking for.”

Daley says it’s great for parents to encourage their children to interact with technology, but they need to supervise their activity.

“It’s also important, like any social media use, any computer use, that you monitor what’s happening and that you have open and frank conversations about the ways the technology can go wrong.”

In July, in response to an update, Grok started making violent, sexual threats on X and dubbing itself “MechaHitler.”

xAI apologized for the situation and said it had been fixed, but some experts question whether the technology lacks sufficient protections.

“Some companies have very strict guardrails because you don’t know who is on the other side of the keyboard,” said Daley. “You don’t know who’s interacting with that, what their social context is. It could be a child, it could be someone experiencing a mental health crisis.”

While he says there is likely a demographic that believes in absolute free speech, he believes the average user tends to want some guardrails.

Countless videos online show just how unpredictable Grok can get in the NSFW setting — particularly in the “unhinged” mode. 

xAI writes that instructing the chatbot to be unhinged “may result in Grok responding like an amateur stand-up comic who is still learning the craft — sometimes being objectionable, inappropriate, and offensive.”  

Videos posted to social media by Tesla drivers using the “unhinged mode” show just how “inappropriate” the chatbot can get using words like c**t and the N-word.

Nasser says she’s hopeful about the potential of AI, but it needs proper protections.

“I love AI. I use it for all kinds of things,” she said.

“But I think we have to think about what we learned with technologies like cell phones, with technologies like social media … and see the lessons that we learned and really apply them to this new wave, this new AI revolution.”