Sunidhi Ahuja
June 1, 2026
In December 2025, a UAE court made history. A law firm was ordered to pay AED 282,508, not for losing a case, not for giving bad advice, but for submitting fake court cases that an AI tool had invented and that nobody bothered to verify.
This is not a story about artificial intelligence being dangerous. It is a story about what happens when professionals forget that they not their tools, are responsible for what they sign their name to.
If you are a lawyer, business owner, compliance officer, HR director, or anyone who has ever let an AI tool draft something important, this ruling is for you.
The case is Arabyads Holding Limited v. Gulrez Alam Marghoob Alam [2025] ADGMCFI 0032, decided by the Abu Dhabi Global Market (ADGM) Court of First Instance on 18 December 2025.
The background: Arabyads Holding Limited brought a commercial claim against a former employee, alleging breach of a share option plan worth over USD 11.2 million. The defendant’s legal representatives, MIO Legal Consultants LLP, filed a defence in response. That defence ran to 327 paragraphs and 233 pages of exhibits.
When Arabyads’ lawyers sat down to review it, they discovered something extraordinary:
several of the legal cases cited in the defence did not exist. Others existed but were cited for propositions they did not actually support. The defence had been prepared using an AI tool Gemini, Google’s AI assistant and the AI had done what AI tools sometimes do: it hallucinated. It invented cases, fabricated citations, and presented them with complete confidence.
The law firm filed them anyway, without checking.
Justice Paul Heath KC presided over the costs judgment. His ruling did two things that matter enormously for anyone working in the UAE.
First, he ordered MIO Legal Consultants LLP to pay AED 282,508 to Arabyads on an indemnity basis, the highest and most punitive cost order available under ADGM rules. This means Arabyads was entitled to recover virtually all of its legal costs incurred in dealing with MIO’s fictitious, bloated defence.
Second, and more importantly, he set out the standard by which all lawyers, and by extension, all professionals, should now be operating when using AI.
His exact words:
“The fault for reliance on AI hallucinations as factually accurate lies not with the research programme, but with the person responsible for conducting the search.”
He went further. The court held that lawyers using AI for research must start from the premise that any authority or article the AI produces may not exist, or may not say what the AI claims it says. The onus is entirely on the professional to verify.
Critically, the court also made clear that intention is irrelevant. MIO Legal did not set out to deceive the court. It did not matter. The professional duty exists regardless of motive.
An AI hallucination is when a large language model, such as ChatGPT, Gemini, or Claude, generates information that sounds authoritative and plausible but is entirely fabricated.
In a legal context, this typically looks like:
A case citation with a real-sounding name, a real-sounding court, and a real-sounding date, that does not existA real case cited for a legal principle it does not actually establish
A statute or regulation quoted with invented wording
These hallucinations happen because AI models are trained to produce fluent, confident text, not to verify facts. They do not search a legal database the way LexisNexis or Westlaw does. They generate plausible-sounding outputs based on patterns in their training data. When they do not know an answer, they often do not say so. They guess convincingly.
This is not a flaw that will be entirely eliminated by better AI. It is a structural characteristic of how these models work. Professional verification is not a workaround. It is the only safeguard.
The UAE Is Not the First — But This Ruling Goes Further
AI hallucination cases have been appearing in courts around the world since 2023. More than 160 have now been documented globally.
The most famous early case involved Steven Schwartz, a New York attorney with 32 years of experience, who submitted six ChatGPT-generated fake cases to a federal court in a personal injury matter. He was fined $5,000 and his career was permanently marked. His defence — that he had no idea ChatGPT could fabricate cases — was not accepted.
UK courts have sanctioned barristers for similar conduct. Qatar’s courts have addressed the issue. Cases have emerged across civil, criminal, and commercial jurisdictions.
What makes the ADGM ruling stand out is its clarity. Justice Heath’s judgment reads less like a penalty notice and more like a professional guidance document. It tells lawyers exactly what the standard is, why it exists, and what the consequence of ignoring it will be. For a jurisdiction like the ADGM, which is closely watched across the Gulf and beyond, this judgment carries significant weight.
Here is where this story becomes relevant to almost every professional in the UAE.
AI hallucinations do not only happen in court documents. They happen in:
The ADGM’s ruling establishes a principle that courts and regulators across the UAE are likely to follow: the professional who submits the document is responsible for what is in it. The AI is not a co-signatory. You are.
If you are a lawyer or legal consultant: Document your AI verification process. Every authority cited in a court filing or legal opinion should have a corresponding verification note. If you cannot show you checked it, assume you will be treated as if you did not.
If you are a business owner or CEO: When you receive legal advice or regulatory guidance that has been prepared with AI assistance, you are entitled to ask how it was verified. If your external counsel cannot answer that question clearly, it is a red flag.
If you are a compliance or risk officer: Any submission to a regulator, court, or arbitral tribunal that contains legal references should pass through a human verification checkpoint before filing. This is now a professional standard, not a best practice.
If you are in HR: Employment policies, termination letters, and grievance procedures that cite legislation or case law should be verified against the actual text of the law. An AI tool that confidently cites a ministerial decision that does not exist has just created a liability.
If you are a finance professional or auditor: Accounting standards, regulatory guidance, and legal opinions that feed into your work need to be verified at source. AI-generated summaries of financial regulations have the same hallucination risk as AI-generated legal research.
The ADGM ruling is particularly significant because of where it sits. The ADGM and DIFC are common law jurisdictions within the UAE, meaning they operate under a different legal framework from mainland UAE courts. They are modelled on English common law, and their judgments are closely read by international law firms, financial institutions, and multinational corporations operating in the region.
Justice Heath’s judgment specifically noted that the hallucination risk is arguably greater in offshore jurisdictions like ADGM and DIFC, precisely because their legal systems draw on a blend of common law and their own bespoke legislation — territory where AI models are even less likely to have accurate training data.
If you are conducting any legal matter within ADGM or DIFC, or instructing lawyers who are, this ruling is the new baseline.
Read More: DIFC in 2026: A Convergence of Capital, Governance, and Modern Dispute Resolution
The ADGM ruling sets a practical minimum standard for anyone using AI in a professional context in the UAE.
Here is what ASK Consultancy recommends:
Verify before you file. Every legal authority, case citation, or regulatory reference generated by an AI tool must be independently confirmed against a primary source before submission to any court, tribunal, or regulator.
Create a verification trail. Document the steps you took to confirm AI-generated content. If a dispute arises, showing your verification process is your first line of defence.
Adopt an AI use policy. If your firm or company does not have a formal policy governing AI use in professional work, the ADGM’s judgment is a reasonable framework for building one.
Assume the AI is wrong until you prove it right. This is not pessimism — it is the standard the court has set. Start from the premise that an AI-generated authority may not exist, and work from there.
Know your exposure in ADGM and DIFC. If you are a party to any proceeding in these jurisdictions and suspect the opposing side has filed AI-generated material without verification, raise it early. The court has shown it will act — and indemnity costs orders are the result.
“AI can research faster than any paralegal. It can also lie more convincingly than any witness. The difference is that a witness faces cross-examination. Your AI-generated brief does not, until opposing counsel reads it.” ASK Consultancy Legal Team
A 327-paragraph defence. 233 pages. AED 282,508. One AI tool that nobody thought to double-check.
The ADGM’s ruling in Arabyads v. Alam is not a warning against using AI. It is a warning against using AI without judgment. The professionals who will thrive in an AI-assisted world are not those who avoid the tools, they are those who understand what the tools cannot do, and make sure a trained human mind fills that gap.
In the UAE’s rapidly evolving legal landscape, that distinction now has a price tag attached to it.
This article is published by ASK Consultancy for informational purposes only and does not constitute legal advice. For specific legal guidance on AI governance, professional liability, or ADGM/DIFC proceedings, contact the ASK Consultancy legal team.
The ASK Consultancy offers expert legal advice in the UAE, specializing in corporate law, residency, citizenship by investment, and real estate. We help clients navigate UAE laws with confidence and compliance. We are also officially registered with DIFC, ensuring adherence to the highest regulatory standards.
Copyright © 2024 ASK Consultancy
