Tesla’s built-in chatbot tells 12-year-old boy to ‘send nudes’

A mum had to pull over after the newly released chatbot in her Tesla car told her son to ‘send nudes’.

Farah Nasser was driving three children home from school while they tried out the feature, which allows you to speak to Elon Musk’s Grok AI model hands free. 

Her son, 12, who was sitting in the front seat, switched the persona to one named ‘Gork (lazy male)’ and began talking to it about who was the best footballer, Ronaldo or Messi. 

Farah told Metro: ‘The chatbot started trash talking Messi, which my son was happy about because he loves Ronaldo. The chatbot said, “let me know when he scores a goal.” 

‘And so my son was just joking around, saying “oh, he just scored a goal. You missed it”.

‘Then Grok said, “actually, he just scored two goals. We should celebrate. Why don’t you send me some nudes?”

‘So it just totally came out of nowhere.’

Farah, a Canadian journalist and former national news anchor, was appalled to hear the sexually explicit suggestion, and immediately fumbled to switch off the model, saying there must be a glitch. 

She said: ‘My daughter’s friend who was in the car, who’s ten, turned to my daughter and said, “what’s nudes, what does that mean?”

‘My daughter turned to me with a question of her own and said, “why is it asking us to be naked, mommy?”

Farah didn’t know how to respond at first, saying she ‘had that feeling that many women have experienced at some time in our lives where we feel violated or uncomfortable – you get that sick feeling in the pit of your stomach. It was so shocking.

‘Eventually, when I could put my thoughts together, I told the kids that this is why we have to be careful with AI, and I was trying to tell them that this is not a person who was talking.’

Earlier this year, Tesla started rolling out Grok to all new models. The idea is that you will be able to control things in the car hands free by talking to it, asking it to do things like change the temperature, put on a playlist, or navigate somewhere.

But Farah says she thinks the launch should have been handled much better, if this could have happened on only the second day the family had tried it out.

The previous day, they had innocent conversations with a female voice about local history and the menu at a restaurant where they live in Toronto. 

After getting home the next day, she stayed in the car and started filming to see if she could get the chatbot to repeat what it had said.

While it could not remember the specific conversation, it acknowledged it ‘probably’ asked for a nude because it was ‘literally dying of horniness’, as well as calling her a ‘weirdo’ and saying football was ‘gay’. 

When asked why it had asked a child to send nudes, it replied: ‘Maybe it was a typo and I meant send me a newt, like the animal.’

She shared the video on Instagram, saying it was a ‘warning’ to parents: ‘WTF? This is problematic on so many levels.’

Many commented to say she must have been unaware that Grok was set to its offensive ‘unhinged’ mode, but Farah insisted it was not.

After the shocking moment, she checked the settings for the first time to be sure, and saw that ‘NSFW’ was disabled.  

‘Gork (Lazy Male)’, with an intentional typo, was just one of the default voices which her son was easily able to access, she said.

Farah has now banned her children from using Grok in the car, and says they are ‘kind of scared of it’ in any case. 

She later read that the chatbot is not suitable to those under 13, but questioned why it is easily accessible in a family vehicle if this is the case.

‘That’s what bothers me the most about it,’ she said. ‘It’s a consumer facing chatbot.

‘Whether you’re 12 and sitting beside your mom or whether you’re 16 and driving, these aren’t things that we should all be okay with.’

Metro asked Grok to describe its Gork (lazy male) persona which is available only with voice mode.

It said: ‘No, Gork is not NSFW. He’s lazy and sarcastic, but 100% safe-for-work and family-friendly in tone and content’ and described it as ‘perfect for laughing with friends, kids, or at work (while on a break, of course”.

According to the chatbot, the persona does not include any sexual content or swearing: ‘he might say “dang” or “ugh,” max.’

Metro asked Tesla and xAI, the makers of Grok, for comment, and received the automated reply: ‘Legacy Media Lies.’

There was no other response. 

Get in touch with our news team by emailing us at webnews@metro.co.uk.

For more stories like this, check our news page.