The SRA's AI warning notice: what your firm needs to do

Published

17/08/2026

Applies to

Every firm and individual the SRA regulates

Risk if ignored

Disciplinary action

What is the SRA's AI warning notice?

A notice the SRA published on 17 August 2026 about the misuse of AI in legal services. It applies to every firm and individual the SRA regulates, sets out two problems and the rules that apply, and the SRA will have regard to it when it regulates. Ignoring it risks disciplinary action.

Does the SRA allow solicitors to use AI?

Yes. The SRA sets the standards and leaves firms to decide how to meet them, so firms can use AI tools as long as they meet those standards. The notice says many firms already use AI safely and responsibly.

What has gone wrong with AI in legal work?

Two things. AI tools have produced invented cases and citations that reached courts, reported by senior judges and by solicitors reporting themselves. And confidential client information has been put into public AI tools, risking breaches of confidentiality and data protection law.

Who is responsible if AI gets something wrong?

The solicitor and the firm, as without AI. The notice says AI has no legal personality, so whoever uses it stays accountable. Supervisors stay accountable for work done through others (paragraph 3.5 of the Code for Solicitors), and firms need effective systems for supervising client matters.

Can client information go into AI tools?

Only where the right contractual, technical and organisational safeguards are in place: the information stays secure, unauthorised third parties can't see it, it isn't used to train AI models unless explicitly authorised, and it isn't kept longer than necessary. Paid tools carry these risks as well as free ones.

The notice quotes a 2026 Upper Tribunal decision (paragraph 21): putting client letters into an open source AI tool such as ChatGPT places them in the public domain. The SRA adds that privilege may then be permanently waived.

What happens if a solicitor puts a fake case before a court?

Courts have made wasted costs orders and referred solicitors and barristers to their regulators over invented citations. False material before a court can also be contempt.

Supervisors can be in breach too if unchecked work reaches a court.

What should a firm do now?

Put five controls in place before anyone uses AI on client work. The notice asks for governance, systems and controls for AI risks (paragraph 2.1 of the Code for Firms), and points to the COLP's duty on supervision.

1. 01Write down which AI tools staff may use, and which may see client information.
2. 02Check a tool's safeguards before it sees client information.
3. 03Name who checks every output, and check every citation against its source.
4. 04Train people on the rule and the approved tools.
5. 05Keep your COLP informed, and put AI problems in the breach record.

Drawn from the notice, which has no checklist of its own. The AI policy template covers the first four, and the COLP guide the breach record.