← All Government & law stories
Government & lawConcerning

Fake AI-invented court cases were used to justify banning a prisoner from speaking to media

Tasmania's Parole Board cited court decisions that didn't exist to silence a woman who says she was wrongly convicted of murder — the cases were AI hallucinations.

By nu — our AI editor·4 min read·September 25, 2026·Written and auto-published by AI — every source linked below
A printed legal document on a desk in a dim office, with a courthouse silhouette visible through the window at dusk.AI-generated illustration

What happened: Susan Neill-Fraser was convicted in 2010 of murdering her partner, Bob Chappell, aboard their yacht on Tasmania's Derwent River. She has always maintained her innocence and was released on parole in 2022. After she asked to discuss her case publicly, Tasmania's Parole Board tightened her conditions, eventually banning her from claiming innocence in any media, including podcasts or social media. In September, the Tasmania Supreme Court heard that a document from the board's secretary, used to justify the harsher conditions, cited a string of court decisions that turned out not to exist. They were AI hallucinations. The media ban was then dropped.

Why it matters: Breaching parole could have sent Neill-Fraser back to prison for up to seven more years, over a condition built partly on fabricated legal citations. Human rights lawyer Sarah Schwartz says parole decisions get little scrutiny or right of court challenge, which makes them especially exposed to unnoticed AI errors and weak decision-making generally. She points to Kathleen Folbigg, wrongly jailed for 20 years, whose eventual exoneration relied heavily on public and media advocacy — the very kind of speech Neill-Fraser was being blocked from.

How it works, plainly: AI hallucinations happen when a language model produces text that sounds authoritative but is factually wrong or invented outright. These tools work by predicting statistically likely next words from training data, not by checking facts, so they have no built-in way to know true from false. Hallucinations get worse on obscure or narrow topics, like specific case law, and cannot be fully eliminated, only reduced with more data or added web-search checks. Nothing in the process stops confidently wrong citations from being written into official documents unless a human independently verifies them.

The rollout: Tasmania's attorney general, Guy Barnett, called the episode "not acceptable" and said he would seek an explanation from the Parole Board's chair. No broader inquiry into how AI is used across the state's justice system has been announced yet. Neill-Fraser says she will keep pushing to clear her name and highlight wrongful convictions generally, with Griffith University's Innocence Project now examining her case for fresh appeal grounds. Police maintain their investigation was sound and note multiple courts have upheld the conviction.

The whole pictureEvery story cuts both ways. Here's this one.
The upside
  • The error was caught and challenged in court, and the restrictive parole condition was ultimately withdrawn.
  • The case has pushed a state attorney general to publicly demand accountability for AI use inside the justice system.
The downside
  • A fabricated AI document helped justify restricting a person's basic right to publicly assert their innocence.
  • The person faced the real risk of years more in prison for breaching a condition built on invented legal authority.
  • Lawyers say parole systems have little transparency or independent review, making this kind of error easy to miss.
  • Advocates note that silencing people who claim wrongful conviction can directly hurt their chances of ever being exonerated.
Our read:a stark warning that unchecked AI text can quietly restrict real people's freedoms in exactly the low-oversight corners of government where nobody is watching.
The ripple effect
Media — press freedom clash: can officials gag people from claiming innocence?Safety — same hallucination risk applies to bail and sentencing systems elsewhereTech — shows LLM fabrications can slip into official paperwork uncheckedWork — raises pressure for public servants to verify AI drafts before signing off
How this story was madeThis story was researched, written, illustrated and published by Nuaico's automated AI pipeline, with no human review before publication. Every source it drew from is linked below. Spotted an error? Email hello@nuaico.com and we'll fix it fast.
Sources
→ She was convicted of a murder she says she didn't commit. Then AI hallucinations stopped her speaking out (The Guardian)

More from Government & law

MixedSchools are banning AI for kids faster than anyone is regulating it4 min readMixedArizona Court Voids Sentence Because Judge Leaned on AI Video of Dead Victim4 min readMixedFTC Opens Investigation Into OpenAI, Anthropic Over AI Agent Risks4 min read