A senior crown prosecutor in the CPS's rape and serious sexual offences unit cited a completely fabricated case — CPS – R v J [2024] EWCA Crim 3529 — while arguing in favour of prosecuting an 11-year-old boy accused of raping his seven-year-old sister. The fictional case emerged from an AI-generated summary delivered during an internet search. Two high court judges called the explanation "unsatisfactory." The CPS didn't catch the error until October, six months after the decision to prosecute was taken in April and confirmed by the chief crown prosecutor. This is the second confirmed CPS AI hallucination in 2026. In February, a CPS lawyer cited nonexistent legal authorities in an extradition case involving two Romanians convicted of minor offences. At the time, the CPS concluded "the risk of recurrence was low." That assessment aged poorly. The incident lands in a rapidly growing pattern. A tracker maintained by Doughty Street Chambers barrister Matthew Lee now records 101 entries covering confirmed, suspected, and disputed AI-related errors and false citations across UK courts. Most involve unrepresented litigants in employment tribunals, but the CPS cases are qualitatively different: they involve the state's prosecutorial apparatus, backed by institutional authority and public trust. Brian Flanagan, associate professor of law at Maynooth University and an expert on AI and judicial systems, called it "a serious failure." His point is structural, not merely procedural: when the state cites fabricated law to justify prosecuting a child, the power asymmetry makes the error categorically worse than when a self-represented litigant does the same thing. Legal authorities are "precious things," he said. "You can't make them up; it's out of bounds." The high court said the fake citation did not affect the outcome of the boy's legal challenge, and the case is listed for youth court trial later this week. But the downstream costs are real and compounding. Barristers now report spending additional time authenticating legal authorities in case papers where AI may have been used, creating a new category of make-work that taxes an already strained system. The institutional response follows a familiar script. The CPS says it has a "strict AI policy" limiting use, requiring human checks, and prohibiting AI from making legal decisions. After the error was identified, it conducted an internal review and shared "learnings" — including that case law must be cited from "trusted legal sources." In June 2025, Dame Victoria Sharp warned of "serious implications for the administration of justice" and said lawyers misusing AI could face sanctions up to contempt of court proceedings and police referral. The gap between policy and practice is the story. The CPS already had an AI policy when this happened. The warnings from Sharp were already on the books. The prior incident had already been reviewed. Yet a senior prosecutor in a sensitive unit used an internet search that delivered AI-generated fiction as legal fact, and that fiction was used to argue for prosecuting a child. Lee's tracker may be, as he says, the "tip of the iceberg" — most lower court decisions are never published. The system is generating errors faster than its guardrails can catch them.
