Artificial intelligence is becoming part of everyday legal work, helping professionals draft documents, summarise information, carry out research and manage increasingly large volumes of material.

But for legal professionals, the consequences of getting it wrong can go far beyond an inaccurate answer on a screen.

The Solicitors Regulation Authority issued a fresh warning to the profession on 17 August, highlighting cases involving inaccurate AI-generated information and confidential client information being entered into public AI tools. The regulator has made clear that using artificial intelligence does not transfer responsibility away from the solicitor using it.

It follows a series of court cases in which fabricated authorities and other AI-generated inaccuracies have found their way into legal proceedings.

Keira Stone, CEO of Kingsley Bond Solicitors, says the real question for legal professionals now is where AI’s role should end.

“AI can be extremely useful in legal practice, but convenience is where people get caught out. A polished answer isn’t necessarily an accurate one, and speed doesn’t transfer any of your professional responsibility onto the tool.

“Treat it as an assistant rather than a colleague. It can help pull together a first draft or summarise a stack of documents, but you still need to know what went into it, and you’re the one who has to stand behind whatever comes out.”

Here, Keira shares six pieces that legal professionals should follow to best capitalise on the benefits of AI.

Know what happens to the information you enter

Before any client information goes in, Stone says, you need to know where it’s going and who has access. “If you wouldn’t send it to an unknown third party, don’t assume it’s safe to paste into an AI prompt,” she says. “Confidentiality comes before convenience.”

The SRA warns that free and paid AI tools alike can pose confidentiality risks without proper protections, and the Law Society advises knowing exactly how client data is processed before it reaches a provider.

2. Never assume a legal citation is genuine

The High Court has already made the point directly, finding that lawyers using AI for legal research have a professional duty to check its accuracy against authoritative sources before relying on it in their professional work. “AI can give you a starting point for research, but it cannot be your authority,” Stone says. “It will invent case names and quotes from judgments that don’t exist, and they read as entirely plausible. Always check the source.”

3. Do not mistake fluent writing for legal judgment

Fluency is not legal judgment. The SRA stresses that human oversight and professional judgment remain essential wherever AI is used in legal work.

“A well-written paragraph can create a false sense of confidence,” she says. “It doesn’t understand a client’s circumstances or weigh competing duties the way a qualified lawyer does, however polished it looks. Legal advice means applying the law correctly to the facts in front of you, and recognising the consequences for the client.”

4. Use approved tools rather than whatever is easiest

“An AI policy only works if people follow it,” Stone says. “Staff need something specific: which tools are approved, and what should never go near them, rather than being left to pick whatever’s easiest.”

5. Keep proper supervision in place

Using AI does not remove the normal requirement to supervise legal work. AI-assisted or AI-generated work still needs proper human review, with an authorised individual carrying ultimate responsibility for what is delivered, and that holds whether a solicitor is using the tool directly or signing off work a junior colleague produced with AI’s help.

“AI should not create a gap in the supervision chain,” Stone says. “If you can’t explain how a piece of work was produced, you’re not in a position to sign it off.”

6. Remember that responsibility stays with the lawyer

“The technology will keep changing, and the tools will keep improving,” Stone says. “But the professional obligation doesn’t disappear because software was involved. If your name is on it, you need to be satisfied it’s accurate before it goes out. AI can save time. It cannot take responsibility.”