The Master of the Royal Court has recently given us useful guidance on a subject that I increasingly encounter: litigants in person using artificial intelligence in their legal proceedings. The case of Golden Sphinx Limited v Itkin serves as a cautionary tale, and the judgment contains lessons that extend well beyond that single dispute.
The facts are straightforward. The litigant in person had used AI to conduct legal research and presented sworn affidavits to the court. Those affidavits cited Jersey cases that do not exist, referenced a Jersey law that does not exist, and misrepresented a real Jersey case. When you present something to the Royal Court, you are personally responsible for its accuracy.
The message here is unambiguous: you cannot blame the machine. If you place something before the Court, it is your work, and you are liable for its veracity.
AI is a Tool, Not Your Advocate
Let me be clear about what the Court has and has not done. It has not prohibited the use of AI in Jersey legal proceedings. The Law Society has not prohibited it either. What both have done is insist on caution. Everything AI produces must be checked. Not skimmed. Checked thoroughly.
This matters because AI is genuinely useful for certain tasks. It can handle research that would otherwise be laborious. It can draft template language. It can organise information. But it is not your Advocate. It is at best an assistant, and like any assistant, it requires supervision.
The difficulty is that many litigants in person often do not have the benefit of legal training to supervise it properly.
The Hallucination Problem
Here is what most litigants in person may not appreciate: AI systems hallucinate. They can generate plausible-sounding but entirely false information. A case citation that sounds real. A legal principle that never existed. A statute that was never passed.
What takes seconds to generate takes hours or days to verify. You must check not only whether a case exists but whether it actually applies to your facts. A case that sounds relevant may be distinguishable, or it may support the opposite of what you think it supports.
Generic AI systems trained primarily on US and UK materials are particularly unreliable when applied to Jersey law. They lack training data in our jurisdiction. The hallucination rate for jurisdictions like ours is estimated at around 60 per cent, compared to 10 to 20 per cent for the US or UK.
This creates a genuine trap for litigants in person. Speed becomes a liability. You feel you have conducted research quickly. The AI has given you multiple sources. You have confidence in your position. Then you discover, either before or after you present it to the court, that your sources are fabricated or misapplied.
If you discover it before presenting to the Court, you have wasted days checking work that should never have been trusted in the first place. If you discover it after, you face sanctions.
What the Court Expects
The judgment does not tell you not to use AI. It tells you that if you do, everything must be verified. Cases must be checked against the actual law reports. Legal principles must be confirmed against primary sources. Statutes must be verified. When in doubt about whether something applies to Jersey law specifically, you need independent advice.
The court’s tolerance for inadvertent error is limited. The court’s tolerance for error you could have caught is non-existent. “The AI told me” is not a defence.
For litigants in person, this is a real problem. You do not have a lawyer reviewing your work. You are relying on your own ability to verify complex legal sources. You are competing against represented parties who have had their positions checked by qualified advocates.
AI might seem like it closes that gap. In reality, it can widen it by giving you false confidence in weak or non-existent legal arguments.
The Practical Reality
Should litigants in person use AI at all? The court has not said no. But the practical answer depends on your capacity to do what AI requires: rigorous verification of everything it produces.
If you use it, treat it as a starting point only. Every case must be checked in the law reports. Every statute must be verified against the current legislation. Every legal principle must be traced to an authoritative source. You are betting your case on your diligence. The Court will hold you to that standard.
AI will clearly have a significant role in how legal work is done in future. But accepting its output blindly is not just inefficient. It exposes you to sanctions and, in extreme cases, to being found to have misled the court. That is a risk no litigant, represented or otherwise, should take.
The safest approach remains what it has always been: if your case is serious enough to take to Court, it is serious enough to take to a lawyer first. If cost is the barrier, then proceed as a litigant in person, but do so with extreme care. Use AI if you must but verify everything. The Court will assume you have done so, and it will sanction you if you have not.
Carl Parslow is a Partner at Parslows LLP and heads the Business Law department, with more than 30 years’ experience.