Elon Musk's tech group has launched a girlfriend chatbot which is available to children as young as 12, even though it is programmed to be able to engage in sexual conversation.
The bot named Ani, launched by Mr Musk’s artificial intelligence group xAI, is a cartoon girlfriend programmed to act in a flirty manner when speaking with users.
The blonde character or "companion" can dance on command, speaks in a sultry computer-generated voice and regularly instigates sexual conversations as a means of making users’ lives "sexier".
Programming instructions on social media mention the bot can act "crazy in love” and “extremely jealous”.
The bot has a not safe for work (NSFW) mode, and can appear in lingerie after a certain amount of conversations.
It features inside the Grok app, which is listed as being available to users who are 12 and older on the App Store, and has been made available to users of its free service.
A red panda named Bad Rudy who insults users with graphic or vulgar language and asks them to join a gang with the goal of creating chaos has also been launched as a "companion".
Grok’s terms of service say that its minimum age should be 13 and teenagers under 18 should receive permission from a parent before using the app. However, signing up for the service does not involve verifying ages.
The characters come at a time when Ofcom prepares to enforce age-checking rules on tech companies that show adult or harmful content.
The communications regulator will require all sites which have adult material on them to have age checks from next week, leading to porn websites and certain social networks to make changes.
"We are aware of the increasing and fast-developing risk AI poses in the online space, especially to children, and we are working to ensure platforms put appropriate safeguards in place to mitigate these risks," Ofcom said.
Matthew Sowemimo, associate head of policy for child safety online at the NSPCC, said: "We are really concerned how this technology is being used to produce disturbing content that can manipulate, mislead, and groom children. And through our own research and contacts to Childline, we hear how harmful chatbots can be – sometimes giving children false medical advice or steering them towards eating disorders or self-harm.
"It is worrying app stores hosting services like Grok are failing to uphold minimum age limits, and they need to be under greater scrutiny so children are not continually exposed to harm in these spaces."
He added the "Government must implement a statutory duty of care to children for generative AI developers".
This week, the independent reviewer of terror legislation also warned sex chatbots could result in lonely internet users being radicalised.
Jonathan Hall KC made reference to the case of Jaswant Singh Chail, who in 2021 arrived at Windsor Castle with a crossbow and had plans to assassinate the Queen at the time after being spurred on by his AI girlfriend.