It began with a woman who simply wanted to speak. Susan Neill-Fraser, a name that has been part of Tasmanian legal history for years, had served time for a crime she says she did not commit. She had been released on parole, but a condition attached to that freedom barred her from talking publicly about her own case. For most people, the right to tell their own story might seem basic. For Neill-Fraser, it was a legal battle. And when she challenged that condition in the Supreme Court, something startling came to light: the document justifying the restriction had been drafted with the help of artificial intelligence, and it cited case law that did not exist. These were not real precedents. They were invented by a machine. Justice Stephen Estcourt set aside the Parole Board’s decision, but the damage to public trust had already begun. A tool meant to make work easier had produced a document with the weight of authority but no underlying truth. The question now is not just what happened in this one case, but how many other decisions, affecting real people’s liberty, may have been quietly shaped by software that fabricated its own sources.
The Greens were quick to respond, and their language was blunt. Cassy O’Connor, the Greens leader in Tasmania’s Legislative Council, said the board’s reliance on artificial intelligence was deeply disturbing and bordering on scandalous. “You cannot delegate matters of justice to a robot, to artificial intelligence,” she said. It is hard to argue with that. The parole board makes decisions about people’s freedoms and parole conditions on a near daily basis. These are not abstract administrative matters. They decide whether someone can walk free, whether someone must report to an officer, whether someone is allowed to see their family, work at a job, or speak in public. A mistake in those decisions can mean a person’s life is turned upside down. O’Connor called for a review of previous parole board decisions to see whether what she called “AI slop, fake case law and hallucinations” were contaminating decision-making. She did not assume the problem was limited to one incident. “It may well be an isolated instance. I suspect not,” she said. Her suspicion carries weight because the legal system depends on the idea that decisions are based on real evidence, real law, and real human judgment. When a machine invents a precedent, it is not simply a technical error. It is a betrayal of the fundamental promise that justice will be done according to law, not according to whatever a computer happened to generate.
The deeper worry is what this reveals about the state of Tasmania’s parole system. Sarah Schwartz, the legal director of the Human Rights Law Centre, described the case as exposing serious deficiencies. She pointed out that the Parole Board of Tasmania is a public authority, and it used AI to generate a document that severely restricted her client’s fundamental freedom to speak. That is a chilling thought. Neill-Fraser was not asking for anything extraordinary. She wanted to speak about her own legal case, a case that has drawn national attention and passionate debate. A public body used a tool that produced fake legal citations to justify silencing her. Schwartz was right to be concerned. If a bureaucratic shortcut can strip away a fundamental freedom, then the safeguards meant to protect ordinary people have failed. And the problem may be bigger than anyone knows. O’Connor also noted that the Justice Department had $68 million cut across the forward estimates in the last state budget. She questioned whether the department and the board had developed their own AI guidelines, as the government’s own guidance recommends. That is a fair question. When budgets are squeezed, agencies look for ways to save time and money. Artificial intelligence can seem like an easy fix. But in the justice system, speed and efficiency can never replace accuracy and accountability. A machine can draft a document in seconds, but it cannot understand the weight of a person’s freedom. It cannot feel the gravity of a decision about someone’s life. And when it makes a mistake, it does so with confidence, inventing legal authorities that look real but are entirely false.
Labor MP Dean Winter also added his voice to the calls for a review. His point was sharp and practical. Neill-Fraser had one of the best-resourced legal teams in the state, he said. That is why she was able to discover the fake case law and challenge it successfully. But what about everyone else? What about people who cannot afford expensive lawyers, who do not have the knowledge or resources to examine every citation, who simply trust that the parole board has done its job properly? Winter asked the question that should be on everyone’s mind: “How often has this happened?” It is a devastating question because the answer is unknown. If a powerful legal team can catch an AI-generated falsehood, how many other decisions have been made based on similar falsehoods and gone unnoticed? The parole system is supposed to be a safeguard, a careful process that weighs risk and rehabilitation, public safety and individual rights. But if AI is quietly infiltrating that process, then the entire system becomes unreliable. The people most affected are often the most vulnerable: those without power, without influence, without the means to push back. For them, a parole decision can feel final. They may not have the ability to demand an explanation. They may accept an unjust condition because they do not know that it rests on fabricated law. That is not justice. That is a lottery.
The state government has responded, though perhaps not as strongly as some would like. Attorney-General Guy Barnett said the production of inaccurate information was not acceptable. He said he would write to the board’s chair seeking an explanation about professional standards and would be “seeking assurances around how the board will ensure this doesn’t happen again.” Those are appropriate steps, but they are also cautious. The call from the Greens and Labor was for a review of past parole board decisions, not just an explanation. And there is a meaningful difference. A letter asking for assurances looks forward. A review would look backward, with the goal of uncovering any other decisions that may have been contaminated by AI. Without that review, the public will be left with a nagging doubt. How many people have been denied parole, or given stricter conditions, because a machine hallucinated a case that never existed? How many people have spent longer behind bars, or lost access to their children, or been silenced, because someone trusted a computer instead of the law? These are not theoretical questions. They are questions about real lives. And they deserve real answers, not just letters and assurances. The government’s own guidance apparently recommends that agencies develop AI guidelines, and O’Connor questioned whether that happened. If the Justice Department did not follow its own rules, then the problem is not only a single staff member who used AI. It is a systemic failure of oversight and leadership. The board should have known what tools were being used. It should have verified the legal citations before relying on them. And it should have accepted responsibility immediately, rather than waiting for a Supreme Court decision to expose the issue.
At its heart, this is a story about trust. The legal system asks us to believe in it. We trust that judges will be impartial, that lawyers will be competent, that parole boards will make decisions based on sound evidence and real law. We trust that the documents filed in court mean what they say. When that trust is broken, it is not easily repaired. Artificial intelligence is not going to disappear. It is already changing the way we work, including the way legal work is done. Used carefully, it might help with research, organisation, and drafting. But it cannot replace human judgment. It cannot be allowed to make decisions about some of the most important moments in a person’s life. The lesson from Susan Neill-Fraser’s case is not simply that machines make mistakes. It is that people must remain accountable for the decisions made in their name. A parole board cannot blame a computer. An attorney-general cannot hide behind assurances. The public deserves a thorough, independent review of past decisions, not because one woman was silenced, but because every person who has ever appeared before the parole board deserves to know that their case was decided by human beings, using real law and honest evidence. Anything less is not justice. It is just a gamble with people’s lives. And the stakes are too high to leave to a robot.