First-person account (JURIST commentary): a Canadian convicted in Taiwan who read his court documents only through consumer chatbots says two systems gave opposite readings of an oversight body's letter in 2025; he published the favourable reading, cited it to journalists and drafted a legal submission on it before the agency confirmed in writing that it was wrong

The proceedings. Cline writes that the courts found that in 2021, during a dispute with his landlord, he posted online documents containing her personal data, including their lease, some of which stayed online for months; he was sentenced to six months in prison and the conviction became final at the Supreme Court on 16 October 2024 (he gives the Taichung District Court, Taiwan High Court Taichung Branch and Supreme Court case numbers). He says he could not read the judgments, no one translated them, and he learned what they say only in September 2026. He is not asking anyone to retry the facts.

Chatbot as interpreter. He writes that across four years no document was translated for him, some hearings were held with no interpreter, and at others the sworn interpreter conveyed a fraction of what was said. 'Every summons, ruling, and prosecutor's letter went into a chatbot, and what came back was the only version I had.' Nowhere in the court file, he notes, does it say that the defendant's understanding came from a language model.

The 2025 mistranslation. In 2025 he ran a letter from a Taiwanese oversight body through two systems, which disagreed 'not on nuance but on the central holding'. He believed the one that told him what he had spent four years wanting to hear, published that reading, cited it to journalists and began building a legal submission on it. In October 2025 the agency confirmed in writing that its letter did not say what he had understood; he withdrew the claim, killed the submission before filing and rewrote his public materials so the correction leads. His media-brief page carries a correction notice: two earlier statements (that the charges proceeded 'despite no harm or malice found' and that later official correspondence appeared inconsistent with the judgment) 'were wrong and are withdrawn', and no authority has found him innocent.

His argument. Fluency is decoupled from accuracy, a chatbot gives no uncertainty signal, and divergence between systems is visible only if one happens to run two; he proposes enforcing interpreter rights as AI policy, requiring divergence signalling where a machine translation supports a legal decision, and counting defendants who read their cases through chatbots. He cites a March 2026 National Human Rights Commission report on interpretation gaps.

Limits. Single first-person account by the subject, who discloses his personal interest; the chatbots are not named; his media-brief page identifies the oversight body as the Control Yuan and dates its written confirmation 28 October 2025; the underlying letter and the divergent outputs were not inspected. JURIST edited and published the piece with a disclosure.

A defendant relying on chatbot translation says a wrong but fluent rendering of an official letter led him to publish a false claim about his case, cite it to journalists and draft a legal submission on it, which he then had to retract and correct publicly (his own account in JURIST and on his media-brief page).

No formal outcome relating to the chatbot use. He writes that the Supreme Prosecutors Office declined an extraordinary appeal on 22 December 2025 and a second petition on 18 September 2026, and that he filed a freedom-of-information request on 3 September 2026 for the number of interpreter appointments in Taiwan's courts. His media-brief page (version 2.0, updated 2 August 2026) publishes the correction.