A 29-year-old man has been arrested in Sydney after telling Claude, Anthropic's chatbot, that he wanted to carry out a mass shooting at a childcare centre. According to a court hearing, David Lan Thackray allegedly wrote that he would "go down to the local Westfield right now and shoot up the local daycare". His Claude account was subsequently suspended.
Thackray was arrested on Sunday 18 October, after the conversation was flagged to Interpol. He has been charged with using a carriage service to menace, harass or offend. At Sutherland Local Court on Monday, prosecutors argued he intended to start murdering people in Sydney, pointing to the Bondi Beach attack in which at least 15 people were killed.
His lawyer, Danilo Rajkovic, rejected the argument, saying there was "no weight behind what he was saying". He said Thackray wrote the messages in a fit of stress and frustration after asking the chatbot for help with an ear infection, and that he had not instructed it to threaten anyone. Thackray has lived in Australia for nearly two decades, is an international relations student and has no criminal history.
The court remanded him in custody for a mental health assessment; the matter returns to court once the evaluation is completed.
The case shows how conversations with chatbots can leave the service. Claude is trained to decline requests to plan or incite violence and can share user data with law enforcement, as already happened in a US case where a woman was reported for written threats to the chatbot. In September Anthropic also published a report on misuse of its models, identifying seven areas of flagged activity, from cyber operations to conventional weapons.
No comments yet. Be the first!