Shooting Daycare Children Using AI? But Luckily Caught

In recent months, as artificial intelligence (AI) has become part of everyday life, there have been growing concerns about people thinking of using it to do terrible things.

A recent incident in Australia has made these concerns even greater. A man reportedly talked to an AI chatbot about doing something extremely violent.

David Lan, who lives in Hurstville, Australia, was arrested on 18 October after Interpol raised an alert about his conversations with Claude, an AI chatbot made by Anthropic.

Police say Lan told the chatbot that he would go to the local Westfield shopping centre right away and shoot children at a nearby daycare centre. He also mentioned the Bondi Beach attack. These statements immediately caused serious safety concerns, and the AI alerted police about the threat.

Lan was charged with using the internet to harass or offend people. He was later kept in custody so that his mental health could be checked.

Prosecutors said the possible danger was catastrophic, meaning it could have been extremely serious. They pointed to the disturbing nature of his searches about shooting children at a daycare centre.

However, Lan's lawyer said he was stressed and was only expressing his frustration while using the chatbot's voice feature to seek help for an ear infection. The lawyer also said Lan believed his conversation was private and had no criminal record.

On one hand, it is fortunate that AI can recognize threats that could put people in serious danger. On the other hand, it is deeply upsetting and worrying that people might even think about carrying out such violent acts.