A Modern Courtroom Problem With an Ancient Lesson

There is something almost timeless about what courts are trying to do: separate what is true from what is asserted, and give each party a fair hearing based on reliable material. But modern disputes now arrive carrying a new ingredient, one that can draft, argue, summarise, and cite at speed.

Generative AI.

In Arabyads Holding Limited v Gulrez Alam Marghoob Alam, the ADGM Court of First Instance confronted a problem that is becoming increasingly familiar around the world: legal documents that refer to authorities which either do not exist or do not support the propositions attributed to them, following what the Court described as the inappropriate use of artificial intelligence.

The result was a clear judicial statement of principle, and a significant costs order.

The Court ordered MIO Legal Consultants LLP to pay AED 282,508 to the claimant as wasted costs, assessed on an indemnity basis, in connection with the defence-related costs issues addressed in the judgment.

But the most important aspect of this decision is not the number. It is the message: technology does not dilute responsibility, especially in a courtroom.

What Happened

This was not a judgment about whether AI should be used. It was about what happens when AI is used without adequate verification in documents filed before a court.

The Court dealt with multiple costs applications arising out of earlier hearings. It explained that these matters are often dealt with summarily and without a published judgment. However, in this case, the Court issued a reasoned judgment because it raised an important question:

When should legal representatives be liable for wasted costs arising from inappropriate AI use in preparing court documents?

In the course of considering that question, the Court referred to problems with citations and authorities in the defence, including examples of “authorities” that did not exist and other citations that were incorrect. It also addressed the professional duties owed to the Court and the risks of placing unverified AI output into a formal pleading.

Crucially, the Court made clear that the responsibility lies with the human professional, not the tool.

As the judgment puts it, the fault for reliance on AI “hallucinations” lies not with the programme, but with the person responsible for conducting and verifying the research.

The Principle the Court Was Defending: Trust

Court proceedings run on a fragile but essential currency: trust.

Not trust in the parties, parties often disagree. Not trust in witnesses, witnesses can be mistaken. But trust in the professional duty that when a lawyer puts an authority before the Court, it exists, it is accurate, and it supports the point being made.

The Court’s reasoning reflects a simple idea: if that trust collapses, the system becomes slower, more expensive, and less fair.

Why? Because the other side must now spend time doing work that should never have been necessary: searching for citations that do not exist, correcting misstatements, and responding to arguments built on unreliable foundations.

In this case, the Court accepted that the claimant’s legal representatives spent “considerable and unnecessary time” reviewing the defence and preparing for hearings, and it found that the inappropriate use of AI research triggered the need for that extensive review.

That is the practical heart of “wasted costs”: time and money spent because something was done in a way that should not have happened.

The Court’s View on AI: Not a Ban, A Standard

This decision does not say “do not use AI.” The judgment is more nuanced, and more consequential.

It says, in effect:

If you use AI, you must treat it like a starting point, not a source of truth.

The Court emphasised that lawyers should start from the premise that authorities revealed by AI research may be inaccurately summarised, or may not exist at all. That means the burden shifts to verification:

Confirm the authority exists. Confirm it says what you claim it says.

In other words, the professional duty is “non-delegable.” You can use tools, but you cannot outsource responsibility.

This principle matters well beyond law.

It applies to anyone using AI for outputs that affect other people, financial decisions, compliance conclusions, risk assessments, policy statements, or regulatory submissions. The faster the output, the more tempting it becomes to trust it. The Court’s message is that speed does not excuse error.

Why the Consequences Were Serious

The Court’s findings on conduct are measured and specific.

It stated that it did not find the firm guilty of conduct intended to mislead the Court. At the same time, it concluded that the way the defence was researched and prepared was a deliberate choice, and it found that the failure to verify AI-assisted legal research was reckless and amounted to a breach of the Rules of Conduct.

The Court described the conduct as unreasonable, “if not improper,” and concluded that the threshold was met for a costs order to be made against the firm.

From a litigation perspective, this is significant because it shows where courts are drawing the line:

You may make mistakes. But if the process demonstrates a failure to apply basic verification and professional judgment, courts will treat the resulting damage, delay, confusion, unnecessary cost, as something that warrants sanction.

The AED 282,508 figure is important not just as compensation, but because the Court recognised that a wasted costs order can include a punitive element as well.

Why This Decision Will Matter Across the Region

The ADGM is not a niche courtroom. It is a leading common law jurisdiction within the UAE, and its judgments are read closely by lawyers, compliance teams, and corporate decision makers across the region.

This ruling signals that:

Courts are ready to treat AI as a professional risk issue.

Not a novelty. Not an excuse.

A professional risk issue.

That has immediate implications for:

• law firms and legal departments using AI for research and drafting
• companies using AI to generate compliance statements or risk positions
• financial institutions using AI outputs to support decision making
• any organisation putting AI generated material into formal, high stakes documents

The lesson is not that AI is dangerous. The lesson is that unverified AI is dangerous, especially where other people must rely on what you submit.

Practical Takeaways for Businesses and Professionals

The simplest way to understand the Court’s approach is this:

AI can help you write. It cannot help you avoid responsibility.

For organisations, that means AI use needs governance, not enthusiasm. The most effective safeguards tend to be practical and boring, because boring is what prevents disasters:

Using AI only within a controlled workflow, requiring human verification of any authority or source, maintaining audit trails of what was relied on, and ensuring senior review before filing or submission.

The Court’s message is that verification is not optional. It is the price of credibility.

How Al Kabban & Associates Can Help

At Al Kabban & Associates, we advise businesses and professionals operating in fast evolving risk environments where technology intersects with legal duty, particularly in disputes, compliance, and governance.

Our work in this area includes:

• litigation strategy and risk management in technology driven disputes
• reviewing and strengthening internal governance on AI assisted drafting and research
• advising on professional responsibility and compliance controls for AI use
• assisting corporates where AI generated outputs raise contractual, regulatory, or reputational exposure

If your organisation is already using AI in operational, legal, or compliance contexts, the key is not whether you use it, but whether you can demonstrate a responsible process around it.

Conclusion: The Tool Is New. The Duty Is Not.

Generative AI is changing how knowledge work gets done. But the courtroom has always insisted on something simple: if you put a claim before the Court, it must be real, accurate, and verifiable.

This decision is best read as a warning, and an invitation.

A warning that shortcuts will be punished when they harm the integrity of proceedings.
An invitation to adopt AI responsibly, with governance that is worthy of professional trust.

In the end, the Court is reminding the modern world of an old rule:

You can automate the writing. You cannot automate accountability.

For more information or to schedule a consultation, contact us at +971 4 453 9090 or visit www.alkabban.com. 

You can also follow us on social media for more updates on everything law related in the UAE: @Alkabban_Law

ALSO READ -

When AI Hallucinates the Law: Singapore’s Landmark Ruling and What It Means for Lawyers Everywhere

When AI Breaks the Rules: What Amazon vs Perplexity Reveals About the Next Great Legal Frontier

AI Regulation in the UAE 2025: What Every Business Needs to Know