Claude AI chat sparks police action after Australian man allegedly threatens to kill children
Police in Australia have arrested a man who allegedly threatened to kill children at a daycare centre while talking to Claude, an AI chatbot made by Anthropic. The case has raised questions about how AI companies watch for threats, protect people's private conversations, and share warnings with police.
Australian man arrested after allegedly threatening children at a daycare centre while chatting with Claude AI
A man in Australia has been arrested after allegedly talking about plans to shoot and kill many people at a childcare centre during conversations with Claude, an AI chatbot made by Anthropic. The case has brought new attention to how AI companies deal with violent threats made in private chatbot conversations and how police find out about these threats.
According to a report by the Sydney Morning Herald, police arrested 29-year-old David Lan Thackray in Sydney on October 18 after Interpol, an international police organization, flagged his conversations with Claude. He allegedly threatened to shoot children at a daycare centre near his home in Hurstville, in southern Sydney. Police charged him with using a communication service to threaten, harass, or offend people.
What did the Australian man say to Claude?
According to police documents shown to Sutherland Local Court, Thackray allegedly told Claude that he would “go down to the local Westfield right now and shoot up the local daycare”.
The court was told that he had been using Claude's voice mode, a feature that lets people talk directly to the AI chatbot instead of typing.
Prosecutors argued that Thackray should not be released. They warned that the result could be “catastrophic” if he was released on bail and broke the rules of his release. They said his conversations suggested that he might be a danger to the public.
Lawyer says the threats came from frustration
Thackray's lawyer, Danilo Rajkovic, argued that the alleged threats were not as serious as prosecutors claimed.
He told the court that Thackray had become frustrated while trying to get advice from Claude about an ear infection. The lawyer said “there was no weight behind” what his client had said, meaning he believed Thackray did not seriously intend to carry out the threats.
The court ordered Thackray to stay in custody while waiting for a mental health assessment, which is an examination of someone's mental health. Reports also said that his Claude account had been suspended.
How did Interpol find out about the Claude conversations?
The case has raised questions about the rules AI chatbot companies use to keep people safe and protect their privacy.
Anthropic, the company that makes Claude, has safety systems designed to stop the chatbot from helping people carry out violent acts. Its rules also allow the company to share certain information with police and other law enforcement agencies during emergencies.
But one important question has not been answered. Public reports have not confirmed whether Anthropic itself reported Thackray's conversations or how Interpol first learned about them.
The case has raised questions about how AI companies handle private conversations that contain sensitive information. It also raises questions about when threats made during conversations with AI chatbots might lead to a police investigation.