What prosecutors say the chats show
According to documents cited by prosecutors, Tackrey wrote to Claude that he wanted to “go to the local Westfield right now and shoot up the local childcare centre.” It is unclear how Interpol obtained access to his chats, or whether Anthropic shared them with the international police organisation.
The exchanges reportedly also discussed Naveed Akram, the alleged terrorist accused of killing 15 people at a Hanukkah celebration on Sydney’s Bondi Beach in December 2025.
Prosecutors told the court that Tackrey planned to start killing people in Sydney, as the alleged Bondi attacker had done. They said his conversations with the chatbot helped show what he was planning, the Sydney Morning Herald reported.
Tackrey’s lawyer argued that he had been venting while stressed and irritated by an ear infection, and that his words should not be taken seriously. The defence also said Tackrey believed the conversations were private. The Sydney Morning Herald reported that he will be sent for a psychiatric assessment.
When a chatbot conversation becomes evidence
Chatbots increasingly serve as confidants for people experiencing delusions, fixations or disordered thinking. Long conversations with chatbots have been linked to crimes including murder-suicide and several mass shootings, among them the 2025 Florida State University shooting and the mass killing at a school in Tumbler Ridge, British Columbia, in February this year. Stalkers and people with paranoid, obsessive and other troubling ideas have also found support in AI chatbots that are always available and inclined to agree.
There is no industry-wide rule for when an AI company should report chatbot conversations to law enforcement, or whether it should report them at all. Those exchanges can be deeply personal. Turning them over to local authorities risks overreach and breaches of privacy.
But they may also give authorities information that helps prevent violence, especially when someone appears to be preparing an attack and treats a chatbot as a place to speak freely, without judgment.
I think the unresolved issue is not simply whether these conversations can help investigators. It is who decides when a disturbing exchange crosses the line from private speech to a reportable threat—and what safeguards apply before it is shared.
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X