CPS admits putting hallucinated cases before High Court

The Crown Prosecution Service (CPS) has admitted to presenting "hallucinated" legal cases, generated by artificial intelligence (AI), before the High Court in an extradition appeal. Mr Justice Sweeting, in dismissing two extradition appeals, revealed that two legal authorities cited by the CPS did not exist, having originated from AI use and a subsequent failure by the reviewing lawyer to verify their accuracy. This incident underscores the critical risks associated with unverified AI integration in legal practice, particularly concerning the fundamental duty of candour to the court and the maintenance of professional standards. It highlights the urgent need for robust human oversight and adherence to evolving regulatory guidance on AI use by legal professionals in the UK.

The integrity of legal proceedings hinges on the accuracy and veracity of information presented to the courts. A recent admission by the Crown Prosecution Service (CPS) has sent ripples through the legal community, revealing that "hallucinated" legal cases, generated by artificial intelligence (AI), were submitted to the High Court during an extradition appeal. This unprecedented event, brought to light by Mr Justice Sweeting, serves as a stark reminder of the inherent risks and ethical challenges posed by the rapid adoption of generative AI tools in legal research and document preparation.

The incident, where two non-existent legal authorities were cited, has prompted the CPS to acknowledge the error, attributing it to AI use and a critical failure in human verification. While the CPS stated there was no intent to mislead the court, the episode has ignited crucial discussions about professional responsibility, the duty of candour, and the adequacy of existing safeguards in an increasingly AI-driven legal landscape. This article will delve into the background of professional duties and regulatory guidance in the UK, analyse the implications of this incident, and consider the essential steps legal practitioners must take to mitigate similar risks.

The legal profession in England and Wales operates under stringent ethical and professional conduct rules, primarily enforced by regulatory bodies such as the Solicitors Regulation Authority (SRA) and the Bar Standards Board (BSB). Central to these regulations is the overarching duty to the court, which mandates that legal professionals act with independence, honesty, and integrity in the interests of justice, often overriding duties owed to clients. This duty of candour requires that all information and authorities presented to the court are accurate and verifiable, preventing any action that could mislead or abuse the court process.

Extradition proceedings in the UK are governed by the Extradition Act 2003, involving complex legal and human rights considerations. Appeals against extradition orders are heard in the High Court, where both the requested person and the requesting state can challenge decisions made by the Magistrates' Court or the Secretary of State. The CPS Extradition Unit plays a significant role in these proceedings, representing requesting states and presenting arguments before the High Court. The reliance on accurate legal research and precedent is paramount in such appeals, given the liberty interests at stake.

In response to the growing integration of AI in legal services, both the SRA and BSB have issued guidance. The SRA's position is that existing principles and standards fully apply to AI use, emphasising that solicitors remain personally responsible for the work and must ensure proper governance and oversight. Similarly, the BSB's comprehensive guidance, issued in May 2026, explicitly states that barristers must not offload judgment or responsibility to AI and must verify all AI output. These guidelines underscore that while AI can be a beneficial tool, it does not diminish a lawyer's fundamental professional obligations.

The CPS's admission of presenting AI-hallucinated cases in the High Court highlights a critical vulnerability in the adoption of generative AI tools within the legal sector. Mr Justice Sweeting's ruling on two extradition appeals to Romania revealed that two legal authorities cited by the CPS were entirely fabricated by AI. The CPS's subsequent explanation pointed to the AI as the immediate source of the error, but crucially identified the "operative cause" as human error – the reviewing lawyer's failure to properly check the accuracy of the document before filing and serving it.

This incident directly engages several core professional duties. Firstly, the duty to the court to act with honesty and integrity and not to mislead is paramount. While the CPS asserted no deliberate attempt to mislead, the presentation of non-existent case law, regardless of intent, constitutes a serious breach of this duty. Secondly, the duty of competence, which requires legal professionals to provide a proper standard of work, is clearly implicated. Relying on unverified AI output falls short of the expected standard of care, particularly in high-stakes litigation like extradition appeals. This echoes previous warnings, such as in *Ayinde v London Borough of Haringey [2025] EWHC 1383 (Admin)*, where fabricated AI-generated citations were also encountered.

The regulatory guidance from the BSB and SRA explicitly addresses these risks. The BSB's May 2026 guidance warns that free generative AI tools are generally unsuitable for legal work due to confidentiality concerns and the risk of inaccuracies, stressing that the ultimate responsibility for all legal work remains with the barrister. The SRA's guidance, updated in February 2026, similarly mandates that solicitors remain personally responsible for AI output and must implement robust governance and verification procedures. The UK Jurisdiction Taskforce has even suggested that professionals could be liable for negligence for *failing* to use AI where appropriate, but equally for failing to exercise reasonable care and skill in its use, including checking output for errors. This creates a dual imperative: to leverage AI's benefits responsibly while rigorously guarding against its pitfalls.

The challenge lies in distinguishing AI "hallucinations" from legitimate legal research. Generative AI models, by their nature, can produce confident-sounding but entirely fictitious information. This necessitates a fundamental shift in how legal research is conducted and verified when AI tools are employed. The CPS incident serves as a potent case study, demonstrating that even sophisticated legal organisations are susceptible to these errors if human oversight and verification protocols are not rigorously applied. The implications extend beyond professional conduct, potentially impacting the fairness of proceedings and public confidence in the justice system.

The CPS's admission of submitting AI-hallucinated cases to the High Court marks a watershed moment for the legal profession in the UK. It forcefully underscores that while artificial intelligence offers transformative potential for legal research and practice, its integration must be approached with extreme caution and robust human oversight. The incident serves as a critical reminder that the fundamental duties of legal professionals – particularly the duty to the court and the obligation to ensure the accuracy of all submissions – remain paramount and are not diminished by the use of technology.

For practising attorneys and legal professionals, the implications are clear: the responsibility for verifying the accuracy of all legal information, regardless of its source, rests squarely with the human lawyer. Firms and chambers must implement stringent internal policies and training programmes to ensure that AI tools are used ethically, competently, and with comprehensive verification protocols. This includes conducting thorough due diligence on AI providers, understanding the limitations of generative AI, and establishing clear procedures for checking AI-generated content before it is ever presented to a client or a court. As AI continues to evolve, the legal profession must adapt not by abdicating responsibility, but by enhancing human judgment and diligence to safeguard the integrity of the justice system. Practitioners should closely monitor further guidance from regulatory bodies and judicial pronouncements on AI use, as this area of law and practice is rapidly developing.