Mention artificial intelligence (AI) and legal work and you’re guaranteed to get plenty of groans from those with experience of the two. . While AI has the ability to help with low-risk prep work, when it is used for actual legal work, especially any court-facing material, it can be a fountain of nightmares.
The Republic’s legal system has already had a clear warning in the matter of Guerin vs O’Doherty. In this instance, the Court of Appeal had to deal with AI-based submissions that hallucinated – or invented – legal authorities.
Relying on AI to generate material without properly verifying it is a risky option. Generative AI (GenAI) can produce text that appears legally plausible but with absolutely no foundation.
Dr Hippolyte Lefebvre, assistant professor in management information systems at UCD, says GenAI platforms simply can’t be trusted blindly to deliver accurately. “Even a technically powerful model can produce plausible but unreliable or inappropriate results.”
READ MORE
Lefebvre’s point gets to the heart of the matter. If you include an inaccurate AI answer in a document you file to your boss as an explainer, you might get a dressing down. If you do so in court, you can waste court time and risk sanctions.
That’s not to say AI is completely without purpose in the sector. It’s vital that anyone considering using GenAI in any sector, not just the legal sphere, recognises that it is a helper, not an expert. The Courts Service already advises users to make themselves aware of the limits of such tools and their duties to the court.
The ideal use cases for AI are in sorting material, creating timelines, putting checklists together and summarising background documents for others helping you to read. All of these fall under some form of admin banner.
When you move away from standard admin to anything more complex, you’re risking trouble. This covers the likes of finding case law or citing litigation, writing or preparing any document for the court or, of course, making any legal arguments.
In this regard, Nicola Flannery, partner in technology and transformation, and digital trust and privacy lead at Deloitte, has advice for managers in all professions, not just the law.
“Staff should be supported to treat AI outputs as inputs to their own judgment, not as unquestionable truth.”
AI can be used to help a lawyer think, but it can’t be what the legal practitioner relies upon. That responsibility is always with them, so they should verify everything.
To its credit, the judiciary is acting to ensure barristers and solicitors are far better informed about AI. It is developing a practice note to guide practitioners and parties on AI use in litigation.
This is addressing matters such as what AI can and can’t be used for, what practitioners must check, and what needs to be disclosed. It is also addressing what judges should expect and what happens when hallucinated or invented material is put before a court.
Similar discussions are taking place in Britain, with the Law Society of England and Wales calling for a range of actions to ensure that any use of AI in the profession is fair and transparent.
While declarations may force practitioners to pause before filing AI-assisted work, they’re far from a panacea. Such declarations don’t prove material is correct, and the breadth of any declaration could be a factor. Given that so much day-to-day software has AI built in as standard, the definition of AI assistance needs to be clear so that only relevant contributions are considered.
If this sounds headache inducing, then I refer you back to the top of this piece. There are many reasons why legal experts groan at the mention of AI. It’s not just about inaccurate information being produced in court, but rather how even positive uses must come with guardrails.
For the foreseeable future, any legal practitioner should consider GenAI to be primarily an admin tool at most. When it comes to organising and sorting material that you provide, it’s a tremendous aid. Anything that requires it to think involves far too great a risk.
