Thomas Mitchell

However, in my own firm I have recently noticed a concerning development when it comes to obtaining client instructions.

The Law Society of Scotland is very clear on this point. Solicitors can only work on their client’s behalf when they have their client’s permission.

However, with the advance in AI, clients are turning to AI tools more often and thereafter reverting with queries which have clearly been generated not by them, but by the AI software, churning out a bullet point list of matters to apparently be addressed.

ChatGPT, Claude, Gemini - AI platforms are here to stay, so how do lawyers adapt to this new normal?

Some of us jump in with two feet, wielding AI tools like Excalibur pulled from the stone, but one should proceed with caution. Examples of cases like Sabrena Rodney v Gee’z Micro Bar and Pitstop and others and the earlier case of R on the application of Ayinde v Haringey LBC both involved fake, AI generated legal authorities. 

Known collectively as the “AI hallucination” cases, these examples of generative artificial intelligence being used carelessly is becoming more common and practitioners should be careful to not let AI do their thinking for them.

However, the silver lining is that AI does not have any impact on the existing principles which practitioners must adhere to in respect of their conduct and duties to their clients, the court and the profession more generally. Unfortunately, AI advances have resulted in these principles being temporarily forgotten by a select minority.

The more difficult question the profession must wrestle with is, “how will clients’ use of AI tools be regulated when it comes to the provision of instructions?”.

I have seen it countless times in my own firm; the email comes in from a client striking a more formal tone than usual, often accompanied by bold, titled paragraphed headers. Often there is a list of questions affixed to the bottom, some relevant to the issue at hand, some not.

Fundamentally, these questions have been generated by an AI programme, not the client, so can you progress on the basis of what the AI programme wants as opposed to the client?

More concerning still, is having to debate with clients about the robustness of one’s advice because “my AI programme says you are wrong”. It takes the classic, “my mate down the pub says you should have done this” argument to a whole new level!

Perhaps I’m a Luddite, maybe even a fool, but I yearn for a time before AI. I am not blind to the obvious benefits, but one can’t help but feel a little dejected by this new normal. At the end of the day, AI suffers from the same issue all computer programmes suffer from, the quality of information it is fed has a direct effect on the quality of the information it spits out the other end. This is known in computer science as “garbage in, garbage out” and as long as garbage keeps going in, I suspect I’ll still be in a job for a wee while longer.

Thomas Mitchell is a partner at RTA LAW LLP