By May 2026, a public database run by researcher Damien Charlotin at HEC Paris had logged roughly 1,490 cases in which AI-generated material was submitted to a court – up from about 200 a year earlier, and still climbing by several a day. Can lawyers use ChatGPT without risking their careers?. Behind almost every entry is the same story: a lawyer who trusted the technology, skipped the verification, and watched their credibility collapse in front of a judge. And yet here is the paradox for any Nigerian law firm weighing generative AI. The same tool producing these disasters is also the biggest leap in legal productivity since online databases replaced the physical law library. When used well, it drafts, summarises, and researches in a fraction of the old time. When used carelessly, it is the fastest way to lose your licence. This piece is about staying on the right side of that line.
How Generative AI Helps Law Firms: Drafting, Review and Research
- First Drafting: Generative models can quickly produce a solid initial draft of documents like non-disclosure agreements, demand letters, board resolutions, or leases. This is not the final product, but for a junior associate facing tight deadlines, it is a game-changer.
- Document Review and Summarisation: Input a lengthy 200-page agreement or a stack of discovery documents, and AI tools can identify key themes, flag inconsistencies, and distil everything into a summary that a partner can review. What used to take a weekend of due diligence can now be completed in just an afternoon.
- Legal Research: With some caveats, AI-driven legal tools significantly speed up the research process, helping lawyers locate authority and “helping lawyers find authority quickly, break down unfamiliar areas, and pressure-test arguments.
What Generative AI Can’t Do: Hallucinations Explained
AI does not replace lawyers. Instead, it compresses the time from initial instruction to first draft and shifts how lawyers allocate their time – less on mechanical tasks and more on critical judgment. In a landscape where clients increasingly resist paying high rates for junior-level grunt work, this evolution is not just beneficial but essential for future billing models.
However, there is a critical aspect often glossed over in promotional materials. The efficacy of AI assumes that the information provided is accurate, but that is not always the case. A large language model does not inherently grasp the law; it predicts the next likely word based on patterns from its training data. Ask it for authoritative support on a shaky legal argument, and it may generate fictitious cases, complete with convincing party names, citations, and quotes – yet these cases may never have existed. Lawyers refer to this phenomenon as “hallucination.”
A study from Stanford RegLab highlights that even AI tools specifically designed for legal purposes can produce hallucinations for a notable percentage of queries, which is a stark reminder that “legal-grade” does not mean “verified.”
Why AI Hallucinates More on Weak Cases
Layered over these technical limitations is a psychological trap. Researchers note that when legal arguments are challenging to substantiate, the AI tends to hallucinate more, trying to meet the user’s expectations. Thus, the weaker the case, the more confidently the tool generates fictitious support, essentially automating confirmation bias.
AI Fake Cases in Court: Real Sanctions, Real Consequences
The cautionary tales are no longer rare incidents. For instance, in the case of Mata v. Avianca (2023), a New York lawyer faced sanctions after relying on ChatGPT, which produced six entirely fabricated cases for a brief against an airline. The judge labelled it an “unprecedented circumstance.” Since then, penalties have escalated, revealing a troubling trend.
In Wadsworth v. Walmart (2025), attorneys from one of the largest US firms, Morgan & Morgan, filed a brief citing nine cases – eight of which were fabricated by their own AI tool. The repercussions were severe: a partner was fined and removed, while supervising lawyers who signed off without reviewing the content also faced penalties. The takeaway was clear: signatures carry liability. Each lawyer is accountable for every citation, whether drafted manually or generated by AI.
In Whiting v. City of Athens (2026), the US Sixth Circuit imposed its strictest sanction of $15,000 per attorney for submitting briefs overwhelmed with fake citations. In another instance, a lawyer received a six-month suspension after failing to disclose AI involvement until the last moments of oral arguments. Courts have begun treating cover-ups as worse than the initial errors.
This issue is not confined to the United States; it has global relevance. Similar cases, such as Ko v. Li in Toronto and Ayinde v. Haringey in England, have emerged, illustrating a pattern across common-law jurisdictions. It is only a matter of time before we see reported incidents in Nigeria. The pressing question for us is whether we will learn from the experiences of others or add to the growing list of missteps.
Where Nigerian Law Actually Applies to Generative AI
This is not a lawless frontier. If you are a Nigerian lawyer considering generative AI, remember that you are already bound by at least three regulatory frameworks. Ignoring this could lead to serious professional repercussions.
Duty of Competence and Candour: Your Obligations Do Not Change
Under the Rules of Professional Conduct 2007 (as amended), lawyers must represent clients competently and engage candidly with the court. The fact that a machine drafted a document does not change these obligations. The American Bar Association’s Formal Opinion 512 (July 2024) clearly states that, when using generative AI, a lawyer retains all duties related to competence, candour, and confidentiality. This includes the responsibility to verify the output, which ultimately falls on the lawyer, not the AI. Submitting a fictional case authority to a Nigerian court is not an AI issue; it is a direct violation of your duties to the court.
Confidentiality and the NDPA: Where Client Data Goes
This is where many firms underestimate the risks involved. Have you ever considered where your client’s confidential information goes once you input it into a public AI tool? According to the Nigeria Data Protection Act 2023, any law firm handling personal data acts as a data controller (or processor), and is obligated to adhere to principles of lawfulness, purpose limitation, data minimisation, and, most critically, confidentiality and security. If you share a client’s private information with a chatbot that retains and trains on those inputs, you could unintentionally violate the NDPA and breach your professional duty of confidentiality in one careless action. For data controllers of significant scale, NDPA penalties can reach a hefty ₦10 million or 2% of annual gross revenue. Coupled with the constitutional right to privacy under Section 37 of the 1999 Constitution, the stakes are tremendously high.
Supervision and Liability: Why Partners Carry the Risk
If a partner allows an associate to work unsupervised with an AI tool, that partner assumes the risk. Take heed of the “Wadsworth case”, which showed that liability can flow up the chain to everyone involved. Nigerian legal principles of supervision dictate that a principal cannot simply delegate accountability to a junior associate, or to a machine, for that matter.
6 Steps on How to Build an AI Policy for a Law Firm
Choosing to ignore AI may not be as safe as it seems. Instead, it could lead to the risk of falling behind firms that harness innovation to provide faster, more cost-effective services. The smarter approach is a governed adoption of AI, characterised by a few crucial commitments.
- Establish a Written AI Policy: Create a comprehensive AI policy before any incident occurs. In one US case, a firm faced consequences despite circulating cautionary guidelines; a formal policy must be supported by a culture of compliance.
- Draw Clear Boundaries: Use AI primarily for drafting, summarising, and ideation. Remember, AI-generated output is essentially a draft until a qualified lawyer has independently verified every citation, figure, and factual claim against primary sources.
- Verify everything: This fundamental rule could prevent most issues associated with AI use.
- Choose Enterprise-Grade Tools: Opt for confidentiality-respecting AI solutions rather than consumer chatbots for any work involving client data. Scrutinise data retention and training terms as rigorously as you would any outsourcing contract.
- Train your team: Train your team not merely on how to interact with AI but on its potential shortcomings as well.
- Disclose when in doubt: Keep in mind that more courts are mandating disclosures involving AI.
None of these steps are unconventional. They mirror the principles of defensibility, transparency, and verifiability that have always been hallmarks of responsible lawyering, now adapted to modern tools.
Conclusion
Generative AI is not here to replace lawyers. Instead, lawyers who understand how to leverage AI effectively, within a disciplined framework, will progressively outshine those who ignore or misuse it. This technology favours sound judgement and penalises laziness -principles that have long defined the legal profession.
The 1,490 cautionary tales out there should not discourage AI use; they should serve as a warning against careless application. Firms that grasp this distinction will build a competitive advantage over the next decade, while others may find themselves justifying their choices in a courtroom.
If your organisation is planning to adopt AI and aims to do so responsibly, reach out to us today!
