What does the SRA actually say about lawyers using AI?

There's a myth doing the rounds among busy solicitors: that the SRA has either banned AI outright or quietly given it the all-clear. Neither is true, and the reality is a lot more useful than either rumour.

The SRA's position lines up with how it handles every other piece of technology a firm brings in. Use it if you want, but the professional duties stay with you. The regulator hasn't bothered writing a separate rulebook for AI because it doesn't need one. The principles that already exist, confidentiality, competence, supervision, cover the ground perfectly well.

The duty that does most of the work

Confidentiality is the big one. The SRA's guidance on AI and technology is blunt about it: solicitors cannot safely put confidential, sensitive, or privileged client information into public AI chatbots or co-pilots, ChatGPT included. If a firm runs a commercial AI tool over client information, it has to read the contractual terms closely and satisfy itself the data stays secure. The responsibility never shifts to the software company. If something leaks, the solicitor is the one who answers for it.

Competence is next, and it trips people up because they imagine it means becoming a machine-learning expert. It doesn't. It means understanding the tools you use well enough to use them responsibly. Does this tool train on my inputs? What does it do with what I type? Where can the output go wrong? You need working answers, not a computer science degree.

Then supervision. AI output is work product, and like a junior's draft it needs checking before it goes anywhere. A firm should be able to show which AI tools it uses, what client data those tools touch, and how the outputs get reviewed. None of that is exotic. It's the same oversight you'd apply to any other part of the practice.

What this looks like on a Tuesday afternoon

You don't need a fifty-page AI policy gathering dust on a shared drive. You need to be able to answer four questions if anyone asks:

Which AI tools does the firm use? What client data, if any, do they touch? What do the vendors' data terms actually say? And how are outputs checked before they reach a client or a court?

A solo practitioner who can answer those four is in a stronger position than a big firm with a glossy policy nobody reads. The SRA refreshes its AI and technology guidance from time to time, so the precise wording can shift. Check the current pages before you lean on a specific position. The underlying logic has held steady though. The convenience is yours to use, and the responsibility is yours to carry.

If you want to see which specific tools clear that bar, our Best AI Tools for Lawyers page scores each one against exactly these duties.

FAQ