The SRA’s new warning notice on the misuse of artificial intelligence is not really about some distant future. It is about what is already happening in legal practice. AI tools are being used to summarise documents, draft letters, assist with legal research, prepare chronologies and make routine work quicker. Used sensibly, they can be genuinely helpful. Used carelessly, they can create serious regulatory problems.

Published on 17 August 2026, the warning notice applies to SRA-regulated firms and individuals. The message is straightforward: if a solicitor uses AI, they are still responsible for the work they produce, the advice they give and the confidential information they handle.

This is why the warning matters. False case citations have already reached the courts. Confidential client information has already been put into public AI tools. Regulators, professional bodies and the judiciary are all making the same basic point: AI can support legal work, but it cannot replace professional judgment.

Why the SRA AI warning notice matters

The SRA is not saying that firms should avoid AI. It recognises that many firms are already using it safely and responsibly. The real point is that AI has to be used within the same professional rules that already apply to solicitors: competence, confidentiality, supervision, integrity, accountability and the duty not to mislead.

There are two obvious concerns. First, AI can produce material that looks convincing but is wrong. That includes fabricated cases, inaccurate quotations and legal analysis that sounds confident but does not survive proper checking. Secondly, client confidentiality and privilege can be put at risk if information is entered into AI systems without proper safeguards.

The notice also has a practical significance. The SRA says it will take the warning notice into account when exercising its regulatory functions. If something goes wrong, a firm may need to explain what checks, training and supervision were in place before the issue arose.

AI hallucinations and false legal citations

The best-known AI risk is the so-called “hallucination”. In simple terms, the system produces something that looks authoritative but is wrong. In a legal context, that might mean an invented case, a false citation, a made-up quotation or a confident statement of law that has no proper basis.

The SRA refers to R (Ayinde) v London Borough of Haringey [2025] EWHC 1383 (Admin), where false AI-generated citations were placed before the court. The lesson is simple. If a lawyer puts a case, quotation or legal proposition before a court, they need to be able to stand behind it. Saying that it came from AI is unlikely to help if nobody checked it properly.

For solicitors, this goes back to the basics. Work must be competent. It must be properly supervised. Arguments must be properly founded. Solicitors must not mislead the court, clients or third parties, including by allowing inaccurate material to be used without adequate scrutiny.

Client confidentiality, privilege and AI tools

The other big issue is confidentiality. Solicitors deal with highly sensitive information every day: health records, financial information, business documents, criminal allegations, family disputes, privileged advice and confidential correspondence. Before any of that information is put into an AI tool, the firm needs to know where it is going, who may be able to access it and how it may be used.

The SRA refers to UK v Secretary of State for the Home Department [2026] UKUT 81 (IAC), where the Upper Tribunal commented on risks arising from client documents being entered into an open-source AI tool. The wider point is obvious. It is not enough to ask whether a tool is free or paid for. Firms need to understand whether data is retained, whether it is used for training, where it is processed, what the contractual terms say and what safeguards protect the information.

A sensible rule is this: fee earners should not put confidential client information into AI tools unless the firm has approved the tool for that purpose and confirmed that the right safeguards are in place. That should be a firm decision, not something left to individuals to work out under pressure.

What other legal bodies are saying about AI

The SRA is not saying anything wildly different from other legal bodies. The Law Society’s guidance, Generative AI: the essentials, takes a balanced approach. AI may create real opportunities for solicitors and firms, but it also brings technology, data, ethical and regulatory risks. The point is to make informed choices about when AI is suitable, what safeguards are needed and how outputs will be checked.

The Bar Council has taken a similar view for barristers. Reliable AI tools can be used to support legal services, but barristers are warned about hallucinations, bias, confidentiality, over-reliance and the need to verify outputs. The Bar Standards Board has also explained how existing professional duties apply when barristers use AI and other technologies.

The judiciary’s October 2025 AI guidance makes the same point from the court’s perspective. Judges are warned about hallucinations and confidentiality. They are reminded that they remain personally responsible for material produced in their name. Practitioners should expect courts and tribunals to be increasingly alert to AI-generated errors.

Practical steps for solicitors and law firms

Firms do not need a huge AI governance project to take sensible steps. They do, however, need arrangements that are clear, proportionate and capable of being explained if the SRA, a client, an insurer or a court later asks what controls were in place.

• Adopt a clear AI policy setting out which tools are approved, what they may be used for and what is prohibited.
• Prohibit the use of confidential client information in public AI tools unless the firm has approved a secure process.
• Require lawyers to verify AI-generated cases, citations, quotations and legal propositions against authoritative sources.
• Build AI risks into supervision, especially where work is produced by junior lawyers, consultants, paralegals or support staff.
• Train staff using real examples of hallucinations, confidentiality breaches and privilege risks.
• Review procurement, data protection, information security and contractual safeguards before adopting any AI tool.
• Keep an audit trail showing why a tool was approved, what safeguards apply and how outputs are checked.

These are practical steps, but they also matter from a regulatory perspective. If AI misuse leads to inaccurate advice, misleading submissions, a confidentiality breach or a supervision failure, the SRA is likely to want to know what the firm did to prevent it.

When AI mistakes become SRA problems

AI misuse can come to the SRA’s attention in a number of ways. A judge may refer a matter after false authorities appear in a skeleton argument. A client may complain after receiving inaccurate advice. A firm may need to self-report a confidentiality breach. An opponent may spot fabricated material in correspondence or pleadings. A data incident may show that confidential information has been entered into an unapproved system.

When that happens, the questions are likely to be practical ones. Who used the tool? Was it approved? What information was entered? Was privilege considered? Who checked the output? Who supervised the work? What training had been given? What was done once the problem was identified?

The answers may shape how the issue is viewed. It may be treated as an isolated mistake. It may be seen as a competence issue, a systems failure, a breach of confidentiality, a supervision failure or, in the most serious cases, conduct affecting the administration of justice.

How JMW can help with AI and SRA regulatory risk

JMW’s professional regulation team advises solicitors, law firms and other regulated professionals facing investigation by the SRA and other regulators. We know that firms are under pressure to work efficiently, embrace new technology and control costs, while still meeting their professional obligations.

We can assist with:

• responding to SRA enquiries about alleged AI misuse;
• advising on self-reporting obligations;
• preparing representations where AI-generated material has been submitted to a court or tribunal;
• advising on confidentiality, privilege and data protection issues following AI use;
• reviewing supervision, governance and remediation arrangements;
• supporting COLPs, COFAs, managers and individual solicitors under investigation.

The SRA’s warning notice should not make firms afraid of AI. The better approach is to use it carefully, with proper human oversight. A firm that can show it has thought about governance, training, confidentiality and verification will be in a much stronger position if something goes wrong.

Conclusion

AI is now part of legal practice. That does not change the fundamentals. Solicitors still need to protect client information, check their work, supervise others properly and avoid misleading the court or anyone else.

That is the key point. AI can support legal work, but it cannot owe duties to clients, protect privilege, exercise judgment or take responsibility for the final product. Those obligations remain with the solicitor and the firm.

If a firm or individual may be exposed to AI-related regulatory risk, early advice can make a real difference. The priority is to identify the issue, preserve the evidence, understand the professional duties involved, and respond in a way that is accurate and credible.

Evan Wright is a partner and barrister in JMW’s professional regulation team.