Fake Case Law, Real Gag: When a Parole Board Trusted AI

The Supreme Court of Tasmania declared invalid the parole conditions imposed on Susan Neill-Fraser in December 2024 and May 2025 after the State of Tasmania conceded that the Parole Board of Tasmania relied on a rationale document prepared by its secretary, Ben Hancock, containing legal authorities that the Human Rights Law Centre said were likely AI-generated hallucinations and did not exist. Neill-Fraser, convicted in 2010 of murdering her partner Bob Chappell and released on parole in October 2022, had been barred from asserting her innocence through media, social media, podcasts or third parties, with a breach potentially returning her to prison for the remaining seven years of her sentence; her parole runs until 2032. Lawyer Patrick Coleridge told the court on 15 September 2026 that "those cases didn't exist," and the Guardian reported on 25 September 2026 that the fabricated case law went unnoticed until her lawyers could not find the cited decisions.

Fake Case Law, Real Gag: When a Parole Board Trusted AI On Tuesday 15 September 2026, in a courtroom in Hobart, a lawyer named Patrick Coleridge explained why a 72-year-old woman had spent the best part of two years forbidden from saying she was innocent. The Parole Board of Tasmania had justified the restriction, in part, with a rationale document prepared by its secretary, Ben Hancock. That document relied on legal authorities: earlier decisions, the kind of precedent that tells a decision-maker that what it is about to do has been done before, lawfully, by someone else. “As we now know,” Coleridge told the Supreme Court of Tasmania, “those cases didn't exist.” The parolee was Susan Neill-Fraser, convicted in 2010 of murdering her partner, Bob Chappell, and released on parole in October 2022 after 13 years in prison. The State of Tasmania conceded, and the court declared, that the conditions the board imposed on her in December 2024 and May 2025 were invalid. The state accepted that the board's decision-making had relied on what the Human Rights Law Centre, which brought the challenge, described as “a document containing legal authorities that were likely AI-generated hallucinations and did not in fact exist.” Ten days later, on 25 September, the Guardian set out what that meant in human terms. Neill-Fraser had been barred from asserting her claimed innocence or wrongful conviction through the media, social media, podcasts or third parties. According to the Guardian's reporting, a breach could have sent her back to prison for the remaining seven years of her sentence. Her parole runs until 2032. Under Tasmanian law, if parole is revoked, none of the time spent on parole counts towards the sentence. It is worth stating plainly what the case is and is not about. Neill-Fraser's conviction stands. Every appeal she has brought has failed. Bob Chappell's family lost a husband, father and brother, and they have lived for seventeen years with a public campaign disputing the verdict. None of that is in question here. What is in question is simpler and, for anyone who cares about how states exercise power, more unsettling. A public body used its authority to restrict a citizen's speech about her own conviction. It justified that restriction with case law that a machine had invented. And nobody noticed until her lawyers went looking for the cases and could not find them. Once that happened, the questions multiplied. How many other decisions, taken by bodies that rarely face challenge and whose reasoning rarely sees daylight, rest on the same kind of invented authority? And when the person silenced is someone who says she was wrongly convicted, what happens to the only route by which wrongful convictions have historically been corrected, which is someone refusing to shut up about them? A Missing Man, a Yacht and Seventeen Years of Appeals Bob Chappell disappeared from the couple's yacht, the Four Winds, moored at Hobart, on Australia Day 2009. His body has never been found. Neither has a murder weapon. Neill-Fraser was charged in August 2009 and convicted by a jury in October 2010. She was sentenced to 26 years with an 18-year non-parole period. On appeal in 2012 that was reduced to 23 years, with a non-parole period of 13. From the beginning she maintained that she did not kill him, and a support group formed around her. The most persistent argument concerned a DNA sample found on the deck of the Four Winds. It belonged to Meaghan Vass, a homeless teenager at the time of the disappearance. The prosecution had argued the DNA arrived by secondary transfer, perhaps on the sole of someone's shoe. In 2015 the Tasmanian Parliament passed the Criminal Code Amendment Second or Subsequent Appeal for Fresh and Compelling Evidence Act, creating a new route back to the Court of Criminal Appeal for people who had exhausted the ordinary appeal process. South Australia had created a similar right, under which Henry Keogh's murder conviction was quashed in December 2014 after twenty years in prison on the strength of flawed forensic pathology. In March 2019, Justice Michael Brett granted Neill-Fraser leave to bring a second appeal. He found she had “a reasonable case to present to the Court of Criminal Appeal.” The appeal was heard in March 2021 and went badly for her at a crucial moment. Vass, the witness on whom much depended, told the court she had been aboard the yacht on the night, then recanted, saying she had been pressured. Neill-Fraser's lawyers abandoned reliance on her evidence and pressed the DNA argument instead. On 30 November 2021 the court dismissed the appeal by two votes to one. Justice Helen Wood concluded that the evidence “is not fresh, it is not compelling.” One judge disagreed. Justice Stephen Estcourt said he would “uphold the appeal and quash the appellant's conviction for murder” and order a retrial. Neill-Fraser's final attempt to take the matter to the High Court of Australia also failed. She left Risdon Prison on parole on 4 October 2022, aged 68, having served her full non-parole period. Her daughter, Sarah Bowles, said the family's energy was renewed, “having mum out to clear her name.” That is the background against which the board acted. The parolee was a public figure openly fighting a conviction the courts had repeatedly upheld. For the victim's family, that fight is a source of continuing pain. For her, it is the point of her remaining years. A parole board sits between those positions with enormous discretion and very little scrutiny. How a Media Ban Grew Into a Gag In December 2024 the Parole Board of Tasmania added a condition to Neill-Fraser's parole. She was not to speak “directly or indirectly with any media outlet to claim her alleged innocence and/or wrongful conviction.” In April 2025 the Human Rights Law Centre, a Melbourne-based legal organisation, filed a challenge in the Supreme Court of Tasmania. It argued that the condition was unreasonable, improper and in breach of the implied freedom of political communication that the High Court has read into the Australian Constitution. The board's response, in May 2025, was not to retreat but to widen. The replacement condition barred her from communicating “directly or indirectly”, including through “third parties, written statements, electronic communications, social media platforms, television, radio, podcasts, streaming services, online video-sharing platforms, or any other public or broadcast media”, in order “to assert claims regarding your alleged innocence, wrongful conviction, or dispute the legitimacy of your conviction or sentencing.” The Guardian reported that the board cited the distress caused to Chappell's family. Read that wording slowly. It reached letters, emails, anything posted online and anything said to a third party who might repeat it. It covered not only claims of innocence but any dispute about the sentence. Taken literally, it is hard to see how she could have let a lawyer say anything in public about a fresh appeal. That uncertainty was one of the grounds on which the state eventually conceded the conditions were invalid. In August 2025 Sarah Schwartz, the centre's legal director, set out the principle. “Freedom of speech and political communication are core democratic rights that belong to everyone,” she said, “including and especially people who have been in prison.” The case rolled on into 2026. Then, in August 2026, Neill-Fraser's lawyers raised with the court that the case law referred to in the board's rationale document did not exist. On Tuesday 25 August the board withdrew the condition. Schwartz called it “an important vindication of Susan's rights,” adding that “parole is meant to be about supporting reintegration, not silencing people or excluding them from public debate.” The court still had to rule on the lawfulness of what had been done. On 15 September Justice Estcourt, the same judge who had dissented in 2021, heard the matter. The state conceded that Neill-Fraser had been denied procedural fairness before the conditions were imposed; that the conditions were uncertain, ambiguous and legally unreasonable; and that the decision-making had relied on the document containing authorities that did not exist. The court declared both the December 2024 and May 2025 conditions invalid. Outside, Neill-Fraser said: “I have always maintained my innocence. I am continuing in the quest to clear my name and to advocate for the broader issue of wrongful convictions in Australia.” Notice the sequence. The conditions were live for roughly twenty months, and the challenge took seventeen months to resolve. The restriction fell not because the board reconsidered the principle, nor because a court ruled on free speech, but because lawyers checked the footnotes and found fiction. Had the rationale cited real cases, however weakly, the free speech argument might still be grinding through the courts. The Document That Nobody Checked What is publicly known about the fabricated authorities is limited, and the gaps should not be filled with guesswork. The court heard that Hancock used artificial intelligence to prepare the rationale document, and Coleridge described the references as “AI hallucinations.” No public account has yet identified which AI tool was used, how many authorities were invented, what those phantom cases were supposed to have decided, or whether the board members who imposed the conditions read the document or relied on its summary. The board has declined to say. “The Parole Board does not comment on individual matters,” it told the ABC, “and is unable to provide any further response to the additional questions raised at this time.” The political response was swift. Tasmania's Attorney-General, Guy Barnett, said: “I will be writing to the chair of the Parole Board seeking an explanation about its professional standards after inaccurate information was produced in this case, which is not acceptable.” He added: “I will be seeking assurances around how the Board will ensure this doesn't happen again.” On the evening of Friday 18 September, the Department of Justice confirmed a wider exercise. “The Department will undertake a review to determine the extent to which Artificial Intelligence AI may have been used to inform past Tasmanian Parole Board decisions,” a spokesperson said. That review has critics. Cecily Rosol, the Greens' corrections spokesperson, said of the Department of Justice that “they're effectively investigating themselves,” and called for past board decisions to be re-examined. Her Greens colleague Cassy O'Connor put it more bluntly: “Justice delivered fairly and equitably is too important to be contaminated by AI slop.” Later on 19 September, Barnett went further: “The Parole Board's professional standards will be independently investigated following its use of unverified artificial intelligence in a rationale document.” Details of the investigator and the terms of reference were to be released once finalised; by early October, none had been published. That investigation is independent. The review of past decisions is not. It remains with the Department of Justice, and the Greens' call for every past decision to be re-examined has not been explicitly taken up. The sharpest response came from Greg Barns SC, the barrister who chairs the Prisoners Legal Service in Tasmania. “It's extraordinary,” he told the ABC. He called for anyone involved in generating the fake case law to resign or be sacked. The service announced its own audit, running in parallel with the department's, covering roughly fifty to sixty parole decisions from the past two years. “If case law is used, we're going to check it,” Barns said. He was careful not to pre-empt the result. “I'm not saying it has been,” he told AAP, “but if it has we'll be bringing this to the attention” of government. That caution matters. As of the date of this article, nobody has shown that any other Tasmanian parole decision rests on invented authority. But the reason for the audit is not a suspicion about one secretary. It is structural. The Neill-Fraser rationale was caught because she had specialist lawyers, a live court challenge and a reason to read every citation. Most parolees have none of those things. A Pattern Now Measured in Thousands of Filings Fabricated legal authority is no longer a novelty. It is a genre. Damien Charlotin, a research fellow at HEC Paris, has maintained a public database of legal decisions in which courts and tribunals found, or implied, that a party relied on hallucinated material. When it was last updated on 5 October 2026, it listed 2,149 cases. It only counts cases where a decision-maker addressed the problem explicitly; filings with fake citations that nobody spotted are, by definition, absent. The case that made the problem famous was Mata v Avianca, a personal injury claim against the Colombian airline in the Southern District of New York. The plaintiff's lawyers cited six decisions that ChatGPT had invented, then produced fabricated excerpts when the court asked for copies. On 22 June 2023, Judge P. Kevin Castel fined the lawyers and their firm $5,000. His opinion listed the damage in terms that read, three years later, like a charge sheet for the Tasmanian episode. “Many harms flow from the submission of fake opinions,” he wrote. “The opposing party wastes time and money in exposing the deception.” He added that the practice “promotes cynicism about the legal profession and the American judicial system.” In England, the Divisional Court went further. In June 2025, in a combined judgment on two cases, Ayinde v London Borough of Haringey and Al-Haroun v Qatar National Bank, Dame Victoria Sharp, President of the King's Bench Division, and Mr Justice Johnson set out what the risks of the technology now meant for anyone who relied on it. “Freely available generative artificial intelligence tools, trained on a large language model such as ChatGPT are not capable of conducting reliable legal research,” the court said. “Such tools can produce apparently coherent and plausible responses to prompts, but those coherent and plausible responses may turn out to be entirely incorrect. The responses may make confident assertions that are simply untrue. They may cite sources that do not exist.” The judges warned of “serious implications for the administration of justice and public confidence in the justice system if artificial intelligence is misused.” Australia has its own list. In July 2024, in a family law dispute known as Handa & Mallick, a Victorian solicitor tendered an inaccurate list of authorities generated with AI-assisted software, and was later referred to the Victorian Legal Services Board. In August 2025, Rishi Nathwani KC, a senior defence barrister, apologised to the Supreme Court of Victoria after submissions in a murder trial of a teenager were found to contain fabricated quotations from a speech to the state legislature and non-existent Supreme Court judgments. “We are deeply sorry and embarrassed for what occurred,” he told Justice James Elliott. The judge's reply was a sentence that ought to be printed on the wall of every public body in the country: “It is not acceptable for artificial intelligence to be used unless the product of that use is independently and thoroughly verified.” Until recently, nearly all of these episodes shared one feature. The invention came from a party arguing a case, and the decision-maker caught it. Tasmania inverts that relationship. Here the fabricated authority sat inside the reasoning of the decision-maker itself, and it was the person subject to the decision who caught it. There is precedent for that, too, and it is not reassuring. In 2025 two United States federal judges, Julien Xavier Neals in New Jersey and Henry Wingate in Mississippi, withdrew rulings containing fabricated material. Answering Senator Chuck Grassley, chair of the Senate Judiciary Committee, both blamed staff use of AI. Judge Neals said a law school intern had used ChatGPT without authorisation and against chambers policy. Judge Wingate said a law clerk had used Perplexity, and that the order docketed on 20 July 2025 “was an early draft that had not gone through the standard review process.” The research literature explains why this keeps happening. In a 2024 study in the Journal of Legal Analysis, Matthew Dahl, Varun Magesh, Mirac Suzgun and Daniel Ho, working with Stanford's RegLab, tested general-purpose chatbots with verifiable questions about real federal court cases. They found that legal hallucinations occurred “between 58% of the time with ChatGPT 4 and 88% with Llama 2.” A follow-up study in the Journal of Empirical Legal Studies found that specialist tools sold to lawyers, including Lexis+ AI and Westlaw's AI-Assisted Research, each hallucinated between 17 and 33 per cent of the time. A language model does not look up a case. It predicts what a citation in that position ought to look like. When the case a user wants does not exist, the model can simply produce one, with the right formatting and the right cadence. The only way to find out it is fiction is to go and look. Why Parole Is Where Errors Go to Hide Courts catch fake citations because they are adversarial, public and staffed by people whose job is to check. Parole boards are none of those things by default. They make consequential decisions, largely in private, about people who mostly lack lawyers. The scale is not trivial, even in a small state. The Parole Board of Tasmania's annual report for 2022-23 recorded 198 applications for parole. Seventy-one were granted, 12 refused and 9 withdrawn. A further 71 were adjourned because the applicant had no suitable accommodation to go to. On top of those sit variations of conditions and revocations for people already on parole. Under the Corrections Act 1997, the board is a statutory body appointed by the Governor. It consists of a lawyer and members qualified in fields such as sociology, criminology, penology or medicine. It is not a court. Challenging a parole decision in Australia generally means judicial review in a state Supreme Court. That route is narrow by design. As the Parole Board Queensland explains to prisoners, the court's examination is “limited to the legality of the decision i.e. was there power and was it exercised lawfully , rather than its merits,” and the court “cannot make a new decision in place of the one made by the Board.” In Victoria, the Adult Parole Board states that it “is exempt from the rules of natural justice and its decisions are not subject to appeal or certain forms of review.” The rules of natural justice include the right to see the material the decision-maker relied on. Judicial review also costs money, time and expertise that most people in prison do not have. Where prisoners have been organised and represented, the numbers can change dramatically. In Queensland, in the two years before 2020-21, 33 applications for judicial review of parole decisions were filed. In 2020-21 there were 325, largely because advocates had discovered that litigation was the only way to force the board to decide overdue applications. Debbie Kilroy, founder of Sisters Inside, told the ABC it was “the only way” to get a response. Helen Blaber, director of Queensland's Prisoners Legal Service, noted what that meant for everyone else: “What that meant is everyone else who wasn't filing a judicial review was being shuffled back.” That is the environment in which invented authority can survive. A rationale document goes into a file the prisoner may never see. If they do see it, they are unlikely to have the training to recognise that a neatly formatted citation leads nowhere. The route to challenge it is a Supreme Court action limited to legality. And few parolees attract a human rights organisation willing to litigate for seventeen months. The contrast with how courts now govern themselves is stark. The Supreme Court of New South Wales, under Chief Justice Andrew Bell, issued guidelines for judges that say plainly: “Judges in New South Wales should not use Gen AI in the formulation of reasons for judgment or the assessment or analysis of evidence preparatory to the delivery of reasons for judgment.” Its companion practice note for litigants, in force from February 2025, bans generative AI from producing the content of affidavits and witness statements. Courts have put guardrails around the technology. There is no public evidence that the Parole Board of Tasmania had anything comparable before this case, and it has not said otherwise. So how many other low-visibility decisions might already rest on fabricated authority? The honest answer is that nobody knows, which is precisely the problem. The two Tasmanian audits will give a partial answer for one board over a limited period. No equivalent public exercise is known to be under way for parole authorities elsewhere in Australia, or for the many administrative decision-makers who now have AI tools on their desktops. What the Neill-Fraser case shows is that the error is possible, that it can reach a binding decision affecting a person's liberty, and that the people best placed to detect it are, in most cases, the least equipped to do so. Silence as a Condition of Freedom Parole boards have legitimate reasons to impose conditions on speech in some cases. A parolee can be barred from contacting a victim, from harassing witnesses, from profiting from crime. The interests of victims' families are real, and the distress of relatives who see a conviction disputed in public year after year deserves respect. The question is not whether a board may ever restrict speech. It is whether a board may lawfully forbid a person from saying that she did not do what she was convicted of doing, on pain of being returned to prison. That kind of condition collides with a well-documented bind that researchers call the innocent prisoner's dilemma. Parole systems tend to reward acceptance of responsibility, because remorse and insight are read as signs of reduced risk. People who are genuinely innocent cannot honestly show either. Andrew Malkinson, wrongly convicted in 2004 of a rape in Salford, was given a life sentence with a minimum term of seven years. He served seventeen. His barrister, Edward Henry, later told a court that Malkinson had stayed in prison the extra decade because he would “not falsely confess to abhorrent crimes which he did not commit.” His conviction was quashed in the summer of 2023, after the legal charity APPEAL commissioned DNA testing, having taken up the case after the Criminal Cases Review Commission twice turned it down. Neill-Fraser's case is not Malkinson's. There is no DNA match to a third party, and the courts have repeatedly found against her. But the principle is identical. A person who maintains innocence is making the only claim that could, if true, correct an injustice. A condition that makes that claim a breach of parole does not only restrict one voice. It removes the person with the strongest motive to keep the case alive from the public argument about whether it should be reopened. Schwartz made that point when the condition was withdrawn. “People who have experienced incarceration are often among the most important voices in discussions about the criminal legal system,” she said, “and protecting their ability to speak out and advocate for change is an important safeguard against injustice.” After the September ruling she put it more simply: “A healthy democracy depends on people being able to speak out about matters of public concern, including the justice system itself.” Fresh evidence often surfaces because a journalist or a scientist hears about a case and gets curious, and innocence projects take on cases because the convicted person tells them their story. The Guardian reported that Neill-Fraser had spoken with Dr Robyn Blewer, director of the Griffith University Innocence Project, in late 2024, shortly before the first condition was imposed. It also reported that the project has now taken on her case. Blewer is among those leading the push for an Australian criminal cases review commission. Ahead of a January 2026 seminar in Brisbane featuring exonerees including Kathleen Folbigg, she said: “It's time to move forward. The task in front of us is monumental and we're under no illusion about that.” Under a condition covering any third party and any written statement, even the conversations that begin such work sat in a legal grey zone. Every Famous Exoneration Began as Noise It is easy to treat public campaigning around a conviction as interference, pressure applied from outside to a system that should be left alone. The history of wrongful convictions in the common law world says the opposite. Again and again the formal machinery failed, and noise from outside forced it to work. On 21 November 1974, IRA bombs in two Birmingham pubs killed 21 people. Six Irishmen were convicted. Their appeals failed. What changed their fate was not a sudden change of heart within the system but a sustained campaign. That included World in Action documentaries in 1985 and a 1986 book, Error of Judgment, by the journalist Chris Mullin, later a Labour MP, who tracked down men he identified as the real bombers. On 14 March 1991 the Court of Appeal quashed the convictions. The men had spent sixteen years in prison. Within hours, the Home Secretary, Kenneth Baker, announced a Royal Commission on Criminal Justice. That commission recommended an independent body to investigate possible miscarriages of justice, and the Criminal Cases Review Commission began work in 1997. Seventeen months earlier, on 19 October 1989, the Guildford Four walked free from the Old Bailey after fifteen years in prison for pub bombings they did not commit. Together the two cases shook public confidence enough to force structural reform. The largest miscarriage of justice in British legal history followed the same pattern. Between 1999 and 2015, the Post Office prosecuted more than 700 sub-postmasters on the strength of data from its Horizon accounting system. It was, in the Criminal Cases Review Commission's words, “the most widespread miscarriage of justice the CCRC had ever seen.” In 2021 the Court of Appeal described the prosecutions as “an affront to the conscience of the court.” Yet many convictions remained in place until January 2024, when ITV broadcast Mr Bates vs The Post Office, a drama about the sub-postmaster Alan Bates and the colleagues he organised. Public outrage moved faster than any court. Within days the Prime Minister announced legislation, and the Post Office Horizon System Offences Act 2024, quashing the convictions en masse, came into force on 24 May 2024. In Australia, the defining modern example is Kathleen Folbigg, convicted in 2003 of murdering three of her infant children and the manslaughter of a fourth. In March 2021, ninety scientists and medical experts, including two Nobel laureates, signed a petition to the Governor of New South Wales citing genetic evidence that the children may have died of natural causes. The Australian Academy of Science backed it. A second inquiry followed. Folbigg was pardoned and released on 5 June 2023, after twenty years in prison, and her convictions were quashed by the NSW Court of Criminal Appeal on 14 December 2023. The Academy's conclusion was blunt: “Wrongful convictions will continue to occur without major reforms to justice systems across Australia.” Each of these cases involved someone insisting on innocence, and people outside choosing to listen. Even Tasmania's own 2015 law, the one that gave Neill-Fraser her second appeal, reflects a legislative acceptance that the ordinary appeal system can fail. Nothing in this history proves any particular person innocent. Many campaigns are wrong. Courts have looked hard at the Neill-Fraser evidence and upheld the verdict each time, and a support group's conviction is not evidence. But the lesson of Birmingham, Guildford, Horizon and Folbigg is not that every campaigner is right. It is that the system cannot tell in advance which campaigns are right, and so it depends on the freedom to make them. A rule that silences the person with the strongest claim to be heard does not protect the integrity of the verdict. It removes one of the few mechanisms that has ever tested it. What an Honest System Would Now Have to Check The Neill-Fraser episode ended the way it did because of a chain of fortunate conditions: a determined parolee, a specialist legal centre prepared to litigate for seventeen months, and lawyers who checked whether the cited cases were real. Remove any link and the gag might still be in place, justified by precedents that never existed. Some responses are obvious and cheap. Every citation supporting a decision that affects liberty should be checked against a real law report by a human who signs off on it. Public bodies should disclose when generative AI has helped draft reasons. Rationale documents should be given to the person affected, so they can find errors the body missed. Reviews of past decisions should be independent of the department whose processes failed, and should publish their method and numbers, not just conclusions. The harder response is to accept what Justice Elliott said in Melbourne and apply it well beyond the courtroom. The technology does not know whether a case exists. It knows what cases usually look like. Any institution that lets that output into its reasoning without checking has handed a fraction of its authority to a system that cannot tell law from fiction. Most of the time, nobody will notice. That is not a reason for comfort. It is the risk. Barns has said his service will check the case law in every decision it reviews. Neill-Fraser has said she cannot walk away. Between those two statements sits the question this case has put to every public body that now has an AI tool on its desktop. When a decision about a person's freedom cites authority, does anyone confirm that the authority is real? In Tasmania, for the better part of two years, the answer was no. The only reason anyone found out is that the woman the board had tried to silence went to court rather than stay quiet. References 1. ABC News. “'AI hallucinations' in parole board submission scuttles media ban on convicted murderer Susan Neill-Fraser.” 15 September 2026. https://www.abc.net.au/news/2026-09-15/susan-neill-fraser-media-ban-parole-board-ai-hallucinations/106302950 https://www.abc.net.au/news/2026-09-15/susan-neill-fraser-media-ban-parole-board-ai-hallucinations/106302950 2. Human Rights Law Centre. “Win for Susan Neill-Fraser as parole board gag order found unlawful.” 15 September 2026. https://www.hrlc.org.au/news/susan-neill-fraser-gag-order-unlawful/ https://www.hrlc.org.au/news/susan-neill-fraser-gag-order-unlawful/ 3. The Guardian. “She was convicted of a murder she says she didn't commit. Then AI hallucinations stopped her speaking out.” 25 September 2026. Syndicated at https://www.inkl.com/news/she-was-convicted-of-a-murder-she-says-she-didnt-commit-then-ai-hallucinations-stopped-her-speaking-out https://www.inkl.com/news/she-was-convicted-of-a-murder-she-says-she-didnt-commit-then-ai-hallucinations-stopped-her-speaking-out 4. Supreme Court of Tasmania, Court of Criminal Appeal. Neill-Fraser v Tasmania 2021 TASCCA 12. https://www.supremecourt.tas.gov.au/wp-content/uploads/2021/11/Neill-Fraser-CCA-Composite.pdf https://www.supremecourt.tas.gov.au/wp-content/uploads/2021/11/Neill-Fraser-CCA-Composite.pdf 5. Human Rights Law Centre. “Legal challenge against Tasmanian parole board's new gag order against Susan Neill-Fraser.” 13 August 2025. https://www.hrlc.org.au/news/legal-challenge-susan-neill-fraser/ https://www.hrlc.org.au/news/legal-challenge-susan-neill-fraser/ 6. Human Rights Law Centre. “Tasmanian Parole Board removes condition restricting Susan Neill-Fraser's speech.” 25 August 2026. https://www.hrlc.org.au/news/susan-neill-fraser-speech/ https://www.hrlc.org.au/news/susan-neill-fraser-speech/ 7. ABC News. “Attorney-general demands explanation from Tasmanian parole board over Susan Neill-Fraser AI mistake.” 16 September 2026. https://www.abc.net.au/news/2026-09-16/tasmanian-parole-board-ai-use-susan-neill-fraser-fallout/107156642 https://www.abc.net.au/news/2026-09-16/tasmanian-parole-board-ai-use-susan-neill-fraser-fallout/107156642 8. ABC News. “Tasmania's Justice Department to review AI use in parole board decisions.” 19 September 2026. https://www.abc.net.au/news/2026-09-19/parole-board-ai-use-review-after-neill-fraser-case/107172064 https://www.abc.net.au/news/2026-09-19/parole-board-ai-use-review-after-neill-fraser-case/107172064 9. Pulse Tasmania. “Tasmania's parole board to face review over fake AI case law.” 19 September 2026. https://pulsetasmania.com.au/news/tasmanias-parole-board-to-face-review-over-fake-ai-case-law/ https://pulsetasmania.com.au/news/tasmanias-parole-board-to-face-review-over-fake-ai-case-law/ 10. Ethan James, AAP, via Queanbeyan CityNews. “'Bigger questions' after parole board's AI use blunder.” 18 September 2026. https://citynewsqbn.com.au/2026/bigger-questions-after-parole-boards-ai-use-blunder/ https://citynewsqbn.com.au/2026/bigger-questions-after-parole-boards-ai-use-blunder/ 11. Damien Charlotin. “AI Hallucination Cases Database.” HEC Paris, last updated 5 October 2026. https://www.damiencharlotin.com/hallucinations/ https://www.damiencharlotin.com/hallucinations/ 12. United States District Court, Southern District of New York. Mata v Avianca, Inc., No. 22-cv-1461 PKC , Opinion and Order on Sanctions, Judge P. Kevin Castel, 22 June 2023. https://www.nhd.uscourts.gov/sites/default/files/pdf/Mata-v-Avianca-sanctions-order.PDF https://www.nhd.uscourts.gov/sites/default/files/pdf/Mata-v-Avianca-sanctions-order.PDF 13. High Court of England and Wales, Divisional Court. Ayinde v London Borough of Haringey and Al-Haroun v Qatar National Bank 2025 EWHC 1383 Admin , Dame Victoria Sharp P and Johnson J, 6 June 2025. https://www.judiciary.uk/wp-content/uploads/2025/06/Ayinde-v-London-Borough-of-Haringey-and-Al-Haroun-v-Qatar-National-Bank.pdf https://www.judiciary.uk/wp-content/uploads/2025/06/Ayinde-v-London-Borough-of-Haringey-and-Al-Haroun-v-Qatar-National-Bank.pdf 14. Queensland Law Society Proctor. “Warning about relying on AI.” October 2024 on Handa & Mallick 2024 FedCFamC2F 957 and Dayal 2024 FedCFamC2F 1166 . https://www.qlsproctor.com.au/2024/10/warning-about-relying-on-ai/ https://www.qlsproctor.com.au/2024/10/warning-about-relying-on-ai/ 15. ABC News. “Senior lawyer apologises after filing AI-generated submissions in Victorian murder case.” 15 August 2025. https://www.abc.net.au/news/2025-08-15/victoria-lawyer-apologises-after-ai-generated-submissions/105661208 https://www.abc.net.au/news/2025-08-15/victoria-lawyer-apologises-after-ai-generated-submissions/105661208 16. Josh Blackman, Reason The Volokh Conspiracy . “Two Federal Judges Apologize For Issuing Opinions With AI Hallucinations.” 24 October 2025. https://reason.com/volokh/2025/10/24/two-federal-judges-apologize-for-issuing-opinions-with-ai-hallucinations/ https://reason.com/volokh/2025/10/24/two-federal-judges-apologize-for-issuing-opinions-with-ai-hallucinations/ 17. Matthew Dahl, Varun Magesh, Mirac Suzgun and Daniel E. Ho. “Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models.” Journal of Legal Analysis 16 1 , 2024, pp. 64-93. https://academic.oup.com/jla/article/16/1/64/7699227 https://academic.oup.com/jla/article/16/1/64/7699227 preprint: https://arxiv.org/abs/2401.01301 https://arxiv.org/abs/2401.01301 18. Varun Magesh, Faiz Surani, Matthew Dahl, Mirac Suzgun, Christopher D. Manning and Daniel E. Ho. “Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools.” Journal of Empirical Legal Studies 22, 2025, pp. 216-242. https://onlinelibrary.wiley.com/doi/abs/10.1111/jels.12413 https://onlinelibrary.wiley.com/doi/abs/10.1111/jels.12413 19. Parole Board of Tasmania. Annual Report 2022 to 2023. https://www.justice.tas.gov.au/ data/assets/pdf file/0005/731399/Parole-Board-of-Tasmania-Annual-Report-2022-23.pdf https://www.justice.tas.gov.au/ data/assets/pdf file/0005/731399/Parole-Board-of-Tasmania-Annual-Report-2022-23.pdf 20. Parole Board Queensland. “Judicial Review.” https://pbq.qld.gov.au/parole-in-queensland/judicial-review/ https://pbq.qld.gov.au/parole-in-queensland/judicial-review/ 21. Adult Parole Board of Victoria. “Review of Board decisions.” https://www.adultparoleboard.vic.gov.au/review-board-decisions https://www.adultparoleboard.vic.gov.au/review-board-decisions 22. Supreme Court of New South Wales. Generative AI Practice Note and Judicial Guidelines. 21 November 2024. https://supremecourt.nsw.gov.au/documents/Practice-and-Procedure/Practice-Notes/general/current/PN Generative AI 21112024.pdf https://supremecourt.nsw.gov.au/documents/Practice-and-Procedure/Practice-Notes/general/current/PN Generative AI 21112024.pdf 23. APPEAL. “Andy Malkinson.” https://appeal.org.uk/andy-malkinson/ https://appeal.org.uk/andy-malkinson/ 24. Criminal Cases Review Commission. “Anniversary of the quashed convictions that led to the CCRC's creation.” https://ccrc.gov.uk/news/anniversary-of-the-quashed-convictions-that-led-to-the-ccrcs-creation/ https://ccrc.gov.uk/news/anniversary-of-the-quashed-convictions-that-led-to-the-ccrcs-creation/ 25. Criminal Cases Review Commission. “Post Office 'Horizon' Cases.” https://ccrc.gov.uk/post-office-horizon-cases/ https://ccrc.gov.uk/post-office-horizon-cases/ Tim Green UK-based Systems Theorist & Independent Technology Writer Tim explores the intersections of artificial intelligence, decentralised cognition, and posthuman ethics. His work, published at smarterarticles.co.uk https://smarterarticles.co.uk , challenges dominant narratives of technological progress while proposing interdisciplinary frameworks for collective intelligence and digital stewardship. His writing has been featured on Ground News and shared by independent researchers across both academic and technological communities. ORCID: 0009-0002-0156-9795 https://orcid.org/0009-0002-0156-9795 Email: tim@smarterarticles.co.uk mailto:tim@smarterarticles.co.uk Listen to the free weekly SmarterArticles Podcast https://www.smarterarticles.fm