The Crown Prosecution Service has confirmed that a senior prosecutor cited an imaginary legal precedent in a high court child rape case after relying on an AI search tool. The nonexistent case law was presented to the court before the error was caught, marking what the CPS itself has termed "a serious failure" in professional standards. The incident is not isolated: court filings across the UK now contain dozens of documented AI-related errors, from fabricated citations to invented statutes, as legal professionals increasingly turn to large language models for research without adequate verification.
The prosecutor in question has not been named, and the CPS has declined to specify which AI tool was used or whether disciplinary action is under consideration. What is clear is that the error occurred during a case involving alleged child rape—a context in which procedural integrity is not merely administrative but existential to the defendant's right to a fair trial. The fact that a senior prosecutor, presumably trained in legal research and citation standards, submitted hallucinated case law to a high court suggests either a catastrophic lapse in due diligence or a dangerous overreliance on tools that are known to confabulate with confidence.
The hallucination tax
Large language models do not "search" in any conventional sense; they generate plausible-sounding text based on statistical patterns in their training data. When asked for legal precedent, they will produce case names, citations, and even judicial reasoning that sound entirely credible—and may be entirely fictional. This is not a bug; it is how the technology works. The problem is that the output is formatted with such authority that it can fool even trained professionals, especially under time pressure or when the user lacks domain expertise in the specific area of law.
The UK legal system is now confronting what other professions discovered earlier: AI tools can accelerate research, but they introduce a new category of error that traditional quality control was not designed to catch. A junior associate might once have cited the wrong paragraph of a real case; now, the risk is citing a case that never existed at all. The former is a mistake; the latter borders on fraud, even if unintentional. Courts in the United States have already sanctioned lawyers for submitting AI-generated briefs filled with fake citations, and the UK is now following the same trajectory.
Professional liability in the age of automation
The CPS has acknowledged the error and promised an internal review, but the incident raises a broader question: at what point does reliance on AI cross the line from reasonable efficiency to negligent practice? Legal professionals are bound by codes of conduct that require them to verify their sources and present accurate information to the court. If an AI tool produces a hallucination and the lawyer does not catch it, who bears responsibility—the tool's developer, the lawyer, or the institution that deployed the tool without adequate training?
The answer, in most jurisdictions, is clear: the lawyer. Professional responsibility cannot be outsourced to an algorithm. Yet the proliferation of AI tools in legal practice suggests that many professionals either do not understand this principle or believe the efficiency gains are worth the risk. The result is a slow-motion crisis in which courts are flooded with unreliable filings, judges waste time fact-checking submissions that should have been vetted before filing, and defendants face the possibility of conviction based on evidence or argument that rests on fabricated authority.
Our take
The UK prosecutor who cited an imaginary case did not commit a victimless error. In a criminal trial, especially one involving allegations as serious as child rape, every procedural misstep erodes the legitimacy of the verdict. If the defense can argue that the prosecution's case rested even partly on hallucinated legal authority, the entire proceeding is tainted. The CPS is right to call this a serious failure, but the response cannot stop at internal review. Legal regulators need to issue clear guidance on AI use in court filings, including mandatory verification protocols and disclosure requirements when AI tools are used in research. The alternative is a justice system in which hallucination becomes routine, and the integrity of the law itself is the casualty.
