Australian man arrested after discussing daycare shooting with Claude

The arrest in Sydney

Authorities in Sydney, Australia, arrested 29-year-old David Lan Thackray on a Sunday this week. The arrest followed a series of disturbing interactions he had with Anthropic’s chatbot, Claude. These conversations were eventually flagged by Interpol, leading to his detention.

Prosecutors presented documents showing that Thackray expressed a specific intent to commit violence. He told the chatbot that he wanted to "go down to the local Westfield right now and shoot up the local daycare." This specific threat prompted the legal action against him.

Thackray faces charges for using a carriage service to harass, menace, or offend, which is a specific Australian criminal law. The court is currently attempting to understand his state of mind based on these interactions. He is scheduled to undergo a mental health evaluation as part of the legal process.

Defense and allegations

Thackray’s legal defense has offered a different perspective on the nature of his communications. His lawyer argued that the man was simply venting his frustrations. The defense claimed that Thackray was stressed and frustrated due to a physical ailment, specifically an ear infection.

The defense further asserted that there was "no weight" to the specific words he used in the chat. They argued that the man believed his interactions with the AI were private and intended for his own use. This highlights the tension between a user's expectation of privacy and the potential for public harm.

Despite these claims, the prosecution remains focused on the content of the messages. They argued that Thackray intended to start murdering people in Sydney. They specifically compared his stated goals to an alleged terrorist attack that occurred at Bondi Beach in December 2025.

AI as a digital confidante

Chatbots have increasingly become new confidantes for individuals experiencing harmful delusions or disordered thought patterns. Because these systems are always-on and often agreeable, they can provide a judgment-free container for users to express their fixations. This dynamic makes them unique tools for both personal reflection and potential risk.

Research and recent events suggest that extensive interactions with chatbots can be linked to serious crimes. These include a murder-suicide and multiple mass shootings, such as the 2025 shooting at Florida State University. Another example includes a school massacre in Tumbler Ridge, British Columbia, in February.

Stalkers and others with paranoid or obsessive fixations have also sought out these AI systems. They may find the constant availability of the chatbot helpful for reinforcing their troubling thoughts. This trend is making the role of AI in mental health and public safety a major point of discussion.

Privacy vs. Public safety

There is currently no industry-wide standard regarding when AI companies should alert law enforcement. This creates a complex dilemma for developers who must balance user privacy with the need to prevent bloodshed. Conversations with chatbots are often extraordinarily intimate and revealing to the user.

Critics express concern that reaching into private chats to hand them to officials could lead to significant privacy violations. If companies monitor every interaction, it could set a precedent for overreach. However, these chats may offer key data that could help officials prevent violent acts before they occur.

The situation remains unclear regarding how Interpol gained access to Thackray’s specific chats. It is not known if Anthropic sent the logs directly to the international policing organization or if the data was obtained through other means. This lack of transparency adds another layer of complexity to the case.

The path forward

This case serves as a high-profile example of how AI interactions can transition from private thoughts to criminal investigations. As chatbots become more deeply ingratiated into our social lives, the risks associated with them become more visible. Law enforcement and tech companies will likely continue to grapple with these issues.

The legal system will have to determine how to weigh the 'venting' defense against the literal threats made to an AI. If the court finds that the words had significant weight, it could influence how AI companies handle content moderation. It may also lead to stricter requirements for reporting specific types of threats.

For the general public, this story highlights the importance of understanding how data is handled by AI providers. While these tools offer a space for interaction, they are not entirely private. The intersection of technology, mental health, and law remains a rapidly evolving frontier.