Supreme Court warns lawyers: Verify all AI-generated citations before filing
New Zealand's Supreme Court has issued a direct warning about the risks of using artificial intelligence in court submissions without proper oversight. In Jones v Family Court at WhangΔrei, the court found that a self-represented litigant had filed citations to cases that did not exist - some fabricated entirely by an AI chatbot, others real cases misquoted to support arguments they did not make.
The implications extend far beyond this single case. Courts in the United States, England and Wales, and Australia have documented the same problem. The pattern is consistent across jurisdictions: unsupervised AI use in legal filings can result in financial sanctions, regulatory referrals, and findings of contempt of court.
What the courts are seeing
In the United States, attorneys in Mata v Avianca Inc. submitted ChatGPT-fabricated case citations to federal court and maintained them when challenged. They were sanctioned $5,000 and found to have acted in bad faith. More recently, in April 2026, Sullivan & Cromwell apologised to the United States Bankruptcy Court after filing an emergency motion containing more than 40 inaccuracies, including fabricated citations and non-existent legal sources.
England and Wales courts have reached the same conclusion. In R (Ayinde) v London Borough of Haringey, the English Divisional Court held that freely available generative AI tools "are not capable of conducting reliable legal research" and that lawyers have a professional duty to verify AI-generated research against authoritative sources.
Australia's courts have taken regulatory action. The Victorian Legal Services Board varied a lawyer's practising certificate after he used cases incorrectly, restricting him to practising as an employee solicitor rather than a principal. New South Wales now requires that where AI is used to prepare written submissions, the author must personally verify in the document that all citations, legal authority, and case law references exist, are accurate, and are relevant.
New Zealand's formal guidance
The Courts of New Zealand issued GenAI Guidelines in December 2023 for lawyers, judges, and non-lawyers. The guidance is explicit on key points:
- GenAI chatbots fabricate cases, citations, and quotes that appear to come from real sources, and will confirm false information as accurate if asked.
- Anything entered into a public GenAI chatbot may be retained by the platform and used to respond to other users - confidential, suppressed, or privileged information must never be entered into public AI tools.
- GenAI should be used with caution for legal analysis because it cannot critically examine patterns, does not produce neutral output, and has limited knowledge of New Zealand law and procedure.
The Supreme Court in Jones endorsed these Guidelines and restated a core principle: "You are responsible for ensuring that all information you provide to the court/tribunal is accurate. You must check the accuracy of any information you get from a [generative AI] chatbot before using that information in court/tribunal proceedings."
This is not a one-off ruling. The Employment Court in LMN v STC (No 2) found that a self-represented litigant had cited a case that did not exist and reiterated the verification requirement. The Employment Relations Authority in O'Driscoll v Rehab Co Mobile Limited awarded increased costs against a party who continued to refer to non-existent cases in final submissions despite being directed to the GenAI Guidelines.
When AI can help
Courts have recognised that AI has legitimate uses in litigation. In Ale v Easthope, an Associate Judge commended a self-represented defendant's use of ChatGPT to structure written submissions, noting they were "well structured and easier to comprehend than might have otherwise be the case with a litigant in person." The defendant disclosed her use of AI upfront, which allowed the judge to assess the work accordingly.
AI can assist with translation and document analysis. In VA v JL, ChatGPT-generated translations of chat application logs were accepted as more complete than the original materials.
The common thread: human oversight was maintained throughout. The person using the tool remained responsible for the output.
Practical steps for legal professionals
Anyone using AI in litigation should independently verify every citation and legal proposition against authoritative sources rather than asking the AI tool itself about accuracy. Do not rely on AI uncritically for legal analysis. Be aware that existing professional obligations - including the duty not to mislead the court - apply in full to AI-assisted work.
Protect confidentiality by not entering private, confidential, suppressed, or legally privileged information into public AI tools. Disclose AI use transparently when required - this supports rather than undermines credibility.
Be alert to authenticity issues. The increasing quality of deepfake evidence is a documented concern across jurisdictions. Challenge any potentially incorrect evidence.
For those seeking to understand AI's role in legal practice more broadly, AI for Legal covers the practical applications in legal research, document review, and contract analysis. Paralegals and support staff may find AI Learning Path for Paralegals directly relevant to managing AI-assisted workflows in a law firm setting.
The message from courts across multiple jurisdictions is clear: AI is a tool, not a substitute for professional judgment. The lawyer or legal professional using it remains accountable for what goes into court.
Your membership also unlocks:
