The Law Society of England and Wales has published a non-binding practice note on how solicitors should approach AI in legal practice

Key takeaways

• The note is non-binding. The Law Society says practice notes do not necessarily provide a defence to complaints of misconduct or poor service.
• On AI agents, the Law Society says use "does not transfer or diminish professional responsibility from you or your organisation".
• Entering client data into AI systems without appropriate safeguards could breach confidentiality, and using client information to train AI is likely to require specific client consent.
• On insurance, the note says insurers may be interested in a firm's AI governance and human oversight.

Is the Law Society practice note binding, and which rules govern solicitors' AI use?

No. The note describes itself as the Law Society's view of good practice. Its legal status section says practice notes are not intended to be the only standard of good practice and that "You are not required to follow them but doing so will make it easier to account to oversight bodies for your actions." Practice notes also do not necessarily provide a defence to complaints of misconduct or poor service, the same section says. The note does not say that departing from it is a breach.

Existing ones, according to the note. It says there is currently no AI-specific regulator for legal practice and that AI use by solicitors is governed through existing legal, regulatory and professional obligations, including the SRA Standards and Regulations. AI use may also trigger data protection, equality and consumer protection law, it adds. The note summarises the SRA's August 2026 warning notice as covering two concerns: AI hallucinations that put false information, including citations, into legal work and court submissions, and confidential client information entered into AI systems without adequate safeguards. Our article on the SRA warning notice covers that instrument. A comparable guide from another jurisdiction is summarised in our piece on the NSW Law Society guide, which does not apply in England and Wales.

What does the note say about AI agents?

Responsibility stays with the firm. In the Law Society's words, "The use of AI agents and agentic AI does not transfer or diminish professional responsibility from you or your organisation." The note suggests firms consider whether the autonomy granted suits the task and whether permissions can be limited to those strictly necessary. It suggests restricting access to client information, matter files and internal systems unless there is a clear need and appropriate safeguards are in place. It also suggests human oversight and approval checkpoints before actions with legal, regulatory or client impacts are completed, and clear allocation of responsibility to named individuals. The solicitor remains responsible for verifying final outputs, the note says.

What does it say about client data, consent and privilege?

Safeguards and consent are the two themes. The note says "If you enter client data into AI systems that do not have the appropriate safeguards, your duty to protect confidentiality could be breached." It also says "The use of client information that could breach confidentiality, including to train AI, is likely to require specific client consent." Specific disclosure or consent may not be required for internal administrative uses such as summarising internal materials, according to the note. It advises treating accidental input of confidential information as a potential confidentiality breach. If a firm considers an incident a serious breach of SRA regulatory arrangements, the note says it "must report it to the SRA promptly". On privilege, the note says AI use can affect legal professional privilege where client information is shared with third-party systems, and that there is not yet clear case law dealing directly with privilege and AI in legal practice.

What does it say about governance, charging and insurance?

The note says "Having some form of governance arrangements for AI use is fundamental", and that they should include appropriate safeguards to manage risks, ongoing review of governance and identification of any training needs. The right approach depends on the size and structure of the organisation, it adds. The note lists what arrangements typically include, such as an AI policy, data handling rules, escalation procedures and clear lines of responsibility, and says a firm is not required to have a standalone AI policy. Time spent operating a system, reviewing and verifying output may be charged "where consistent with the client care letter". On insurance, the note says AI use may change a firm's professional indemnity risk profile and suggests discussing any uncertainty with a broker or insurer, particularly at arrangement or renewal. Insurers may be interested in governance processes and in the extent of human oversight of AI-assisted work, it says.

What should solicitors and COLPs do now?

Use the note as a reference for your firm's documented position. The note says that a firm which does not follow a suggested route should be able to justify the alternative to oversight bodies.

• Compare your AI policy, agent permissions and client-care wording with the sections on agentic AI, confidentiality and governance.
• List every AI tool in use, including AI inside software you already license, and note what client data each can reach.
• For any agent that acts on files or systems, record who is responsible and where a human approves before it acts.
• Before a PII renewal, check what the proposal form asks about AI and prepare accurate answers, since insurers may ask.

What are the key dates and parties?

What we did not verify

What we opened: the full practice note page on the Law Society website, including the legal status section, on 24 October 2026.

What we did not open: the SRA warning notice itself (we rely on the Law Society's summary), the court decisions the note cites, the Civil Justice Council consultation, any PDF or versioned copy, and any SRA statement about the note. The date is the page date stamp, and we cite the note by heading because it is a single long page.

What we refuse to claim: that the note is binding, that failing to follow it is a breach, that the SRA has endorsed or adopted it, or what insurers or courts require beyond what the note reports.

Related briefings

• AI Regulation News hub
• Europe AI regulation tracker
• Attorneys and AI regulation
• SRA warning notice on misuse of AI by solicitors
• NSW Law Society guide to responsible AI use for solicitors
• Bar Standards Board guidance on barristers and AI
• ICO statutory code of practice on AI and automated decision-making