As the SRA announces it has approved the first “purely AI-based firm … to provide regulated legal services in England and Wales” it seemed like a good time to take stock of where the AI drive is taking the legal profession, particularly on the Isle of Man.

The SRA licence for the English firm is in respect of recovery of unpaid debts only, and is aimed at guiding the consumer through the small claims process. It seems therefore that its scope is very much restricted at present. The increasing potential for AI in the legal sector is obvious though, and consumers must be tempted to try and use readily available open AI offerings to mitigate the costs of appointing a professional. Similarly, when dealing with Isle of Man legal cases, non Manx lawyers will be tempted to use AI to “Manxify” their work (the Isle of Man being a separate and distinct legal system from that of England and Wales, with its own legislation, rules of court and jurisprudence). All should be cautious to do so however. I hear cries of “bias” so I’ll explain why.

Research

Undertaking research is perhaps the most obvious task which AI could assist with. The problem with asking AI to research an issue however is that:

1. You need to understand the question you are asking to make sure you ask the right thing; and

2. You need some background knowledge of the subject and local law to interpret the answer to make sure it is correct otherwise you place blind faith in the AI tool you are asking.

By way of (a very dry) example, I asked an open AI tool:

“what section of the isle of man companies act 1931 deals with the ability to restore a company post liquidation”

I thought that question was particularly kind to the open AI tool. For a start, I have told it where to start looking for such a provision. It’s not as if I am asking it to consider all Manx law and search to tell me whether restoration of a dissolved company post liquidation is possible but, as to that, then you can’t really liquidate a company if it is dissolved and so restoration would have to take place first.

The answer I was expecting was either, that you cannot liquidate a dissolved company, or I would have expected the open AI tool to point me to s.273(6) Companies Act 1931 (act of Tynwald) dealing with restoration of dissolved companies perhaps. Instead, it told me that s.166 Companies Act 1931 deals with restoration of dissolved companies. It does not. s.166 is about the ability to stay or restrain proceedings against a company which is in liquidation. It was only at the third time of asking that the open AI tool pointed me to the section about restoration. I had to tell it twice that the answer it had given me was wrong (for which it apologised!).

This is not the first time that I have tested open AI tools in respect of Manx law and have been given the wrong answer. There are obvious problems then. OK sure, you can check the AI resource’s answer to make sure it is right but, if you have to do that, you may as well do the research yourself anyway. The real danger is in relying on the first answer it gives you. You might say that there’s a similar danger with lawyers, in response to which I cannot argue that they are infallible. They are required to have PI insurance though which open AI tools are not. There is then some protection for clients.

In general, in testing open AI resources, I have also found them unable to refer me to Manx case law in support of any particular point.

Drafting

If an AI tool cannot properly research, then how can it properly draft?

Even if it could properly research then it does not follow that drafts will be accurate or up to the job. Again the art here is making sure the right question is asked. The problem is that AI will not necessarily ask a potential client the right questions. In drafting a will for example, will an AI tool ask how many dependents there are, what age they should inherit at, what specific gifts (if any) are to be bequeathed, whether there are previous wills, whether the testator has assets in other jurisdictions which may need separate wills or advice, and how will AI assess the capacity of the testator? These are just a few of the questions it would be good to know the answers to in order that a working will can be properly drafted.

Data Security

Take the issue of drafting the will cited above. Let’s assume that the example questions I posed as being relevant were all asked, or that the information was given to the AI tool in the first place. Open AI resources learn from all the data which has been previously input and also from that available on the web. The data which is input is stored by the AI tool to be learned from and is technically available to it ad infinitum. It is therefore stored somewhere and potentially available to the world at large and/or to be stolen or duplicated. Inputting sensitive data about names and addresses, dates of birth, assets and location thereof, perhaps even investment policy numbers, therefore creates an unnecessary risk.

Conclusion

There’s no doubt of the potential for AI to assist us. Its application needs to be regulated and it should be used with caution. AI should be part of the potential solution in provision of legal services, not the solution. There is still no substitute for experience. Neither can AI pick up the ‘phone to another lawyer and have an off the record chat to try and achieve the best outcomes for all parties to a dispute, or to discuss what is relevant in a particular piece of drafting.

At the very least legal firms should have an AI policy for staff members as to use of AI tools. Ideally it would also have a secure AI tool which does not publicly store any client data, so mitigating the risk of data theft or leakage. At M&P Legal we have invested heavily in bespoke and secure AI tools to assist our Advocates in their provision of legal services; not to do the job for them, but to help carry out their instructions more efficiently in the right circumstances.

At M&P Legal, we’re working hard to keep abreast of the latest legal and technological advances to safely incorporate them into the way we work and offer services to clients to maximise efficiency. The author of this article, Damian Molyneux is an advocate and director of M&P Legal. He can be contacted via dpm@mplegal.im. This article does not constitute legal advice and specific advice should be sought for individual circumstances.

Back to top