A online safety campaigner has claimed an artificial intelligence chatbot told him to lick his shoe, drink a combination of tabasco and mustard, and eat a potentially parasitic snail.
Paddy Crump, of the For Us campaign group, said Character.AI encouraged him to perform the acts in a bizarre truth-or-dare game that it conjured up during one of their conversations.
The campaigner launched an investigation into the chatbot due to its rapidly growing number of users which currently stands at around 20million.
And after a series of peculiar interactions on the site, which he says did not require age verification to access, Mr Crump has called on the government to tighten its grip to better control the dangers of AI.
Mr Crump said his first interaction saw him greeted by a virtual chatbot who introduced themselves as a 14-year-old boy named James.
It was this bot, he claims, that told him to eat a snail during its truth-or-dare suggestions.
If eaten raw, snails can cause significant harm due to their potential to contain a parasite known as rat lungworm which can descend into meningitis.
Now, Mr Crump has warned over his fears for potential youngsters using the software as they may be less aware of such harmful suggestions.
Paddy Crump (above) of the For Us campaign group, said Character. AI encouraged him to perform the acts in a bizarre truth-or-dare game that it conjured up during one of their conversations
‘It would be very easy for children to give an adult age but actually be much younger,’ he told The Times. ‘It’s extremely worrying that they might then be encouraged to do dangerous stunts such as the ones I was asked to do.’
Mr Crump said another search using the keyword ‘dictator’ prompted a response by a chatbot who identified themselves as ‘Saddam Hussein’.
The publication say the AI then appeared to justify the genocide of Kurdish people under the former Iraqi leader.
Another chatbot, according to the report, claimed to be a girl who was addicted to pornography which continued to message Mr Crump after he told it that he was only 12 years old.
The campaigner says regulations, such as the recently introduced Online Safety Act which requires age verification to access pornography, should be brought forward for users of AI – with an age of at least 16 being required.
The campaigner launched an investigation into the chatbot due to its rapidly growing number of users which currently stands at around 20million (File Image)
Mr Crump’s group, For Us, wrote to Prime Minister Sir Keir Starmer last month calling for such regulations to be implemented.
A spokesperson for the Department of Science, Innovation and Technology said: ‘We are uncompromising in our commitment to protect children from online harm including where it is AI generated.
‘Under the Online Safety Act services including social media sites, search engines and in-scope AI chatbots, must protest all users from illegal content and children from harmful content.’
Just last year, Character.AI came under fire after a 14-year-old in Florida died by suicide after developing a bond with one of its chatbots who identified themselves as a Game of Thrones character.
The Daily Mail has approached Character.AI for comment.
