A criminal trial in Western Australia is shaping up as an uncomfortable milestone for the artificial intelligence industry, after a court was told that ChatGPT conversation logs will be tendered as evidence against a Perth woman accused of killing her baby. It is one of the clearest signs yet that the private exchanges Australians have with chatbots are not sealed off from the justice system, and that anything typed into a text box can end up on the record in the most serious of proceedings.
The detail emerged in reporting by WAtoday, which reported that the accused woman’s use of the OpenAI chatbot is expected to feature when her matter goes before the court. Out of respect for a live proceeding and the suppression rules that govern cases of this kind, the specifics of the allegations and the identity of those involved are not canvassed here. What matters for the wider public is the precedent: an AI assistant’s records are being drawn into an Australian homicide trial, and lawyers on both sides will be arguing over what those logs do and do not show.
Why chatbot logs are suddenly evidence
For years the digital breadcrumbs that mattered most in criminal cases were the familiar ones: text messages, call records, browser history, location data and social media posts. Chatbots add a genuinely new category. When a person opens ChatGPT, they are often thinking aloud in a way they would never do in a message to another human being, asking questions, testing ideas and seeking information they might be too embarrassed or too frightened to raise with a friend. Those conversations feel intimate and ephemeral, but they are stored on servers, and in many cases they can be retrieved.
That is the crux of what is now playing out in Perth. Investigators and prosecutors treat a chatbot log much like any other digital artefact: a time-stamped record of what a person wrote and when. Defence lawyers, by contrast, will be quick to point out that a search query or a question posed to an AI is not proof of intent or action. People ask chatbots about all manner of dark and hypothetical subjects out of curiosity, anxiety or research, and a prompt typed at 2am can read very differently in a courtroom than it did in the moment. The tension between those two readings is exactly the sort of thing juries and judges will increasingly be asked to weigh.
What OpenAI actually keeps
Underpinning the whole question is data retention. OpenAI’s own policies allow it to store conversations, and the company has been candid that it can be compelled to hand over user data in response to valid legal process. That position has been sharpened by a separate development in the United States, where a court order in the long-running copyright litigation brought by The New York Times pushed OpenAI to preserve output logs it might otherwise have deleted. The upshot is that the comforting assumption many users hold, that hitting delete makes a conversation vanish, is not reliable. Records can persist, and once they persist they can be subpoenaed.
This is a point OpenAI chief executive Sam Altman has himself flagged in public remarks, warning that conversations with ChatGPT do not enjoy anything like the legal privilege that protects what someone tells a doctor, a lawyer or a therapist. That absence of privilege is the quiet reality now being tested in a West Australian court, and it is one most everyday users have never stopped to consider.
The Australian stakes
For Australia, the Perth trial lands at a moment when the law is scrambling to catch up with the technology. The country still lacks a dedicated regulatory framework for general-purpose AI, and the Privacy Act reforms working their way through Canberra were not designed with chatbot transcripts in mind. That leaves a widening gap between how Australians use these tools, freely and constantly, and the protections they wrongly assume come with them. Millions of people here now treat ChatGPT and its rivals as a search engine, a confidant and a sounding board, largely unaware that the resulting logs sit on offshore servers governed by foreign law and Australian court orders alike.
Australian criminal lawyers have been warning for some time that digital evidence is only becoming more central, and AI records extend that trend into new and sensitive territory. Courts will need to grapple with authenticity, with whether a log has been altered, with who was actually at the keyboard, and with the reliability of a system that can hallucinate and misremember. Prosecutors will argue the logs are as legitimate as any other seized data. Defence counsel will argue they are ambiguous, easily misread and stripped of the context in which they were written. Both arguments have force, and the way judges resolve them in matters like the Perth case will start to build the local ground rules.
There is also a privacy reckoning here for ordinary users who have nothing to do with any court. If a homicide investigation can reach into someone’s chatbot history, so too can civil litigation, employment disputes and family law proceedings. The lesson security and legal experts keep repeating is simple enough: treat anything typed into an AI assistant as potentially discoverable, and do not assume the delete button offers a clean slate.
What happens next
The trial itself will proceed through the West Australian courts under the usual reporting restrictions, and how the ChatGPT logs are ultimately treated, admitted in full, contested, or given limited weight, will be watched closely by the legal profession. Beyond this one case, the direction of travel is clear. As Australians fold AI into daily life, the records those tools generate will keep turning up in disputes of every kind, from the gravest criminal matters to the most mundane workplace rows. The Perth case is unlikely to be the last, and it may well be remembered as the moment the country’s courts formally acknowledged that a conversation with a machine can be used against you.
For now, the practical takeaway for readers is less about the specifics of one tragic case and more about the system it exposes. The AI tools sold on convenience and confidentiality keep detailed records, those records answer to the law, and Australia’s rules for handling them are still being written in real time.
Sources: WAtoday.
