On 17 August 2026, the SRA published a warning notice on the misuse of AI in legal services (https://www.sra.org.uk/solicitors/guidance/misuse-ai/). Whilst it still leaves more questions than answers, what it does make clear is that blaming AI is not going to be an acceptable excuse.
Two things the SRA is worried about:
- Hallucinations in court documents. AI tools can confidently generate fake cases, fake citations, and fabricated “facts” that look entirely plausible. The SRA has already received reports from senior judges, plus several self-reports from solicitors who relied on AI-generated content that turned out to be a ‘hallucination’. The misuse of AI tools can result in real harm to clients, the court and to trust and confidence in the legal profession, and those accused of such misuse may face regulatory action, wasted costs order and even contempt of court proceedings.
- Client confidentiality. Entering confidential client information into public AI tools risks breaching confidentiality, and, more seriously, can permanently waive legal professional privilege. Once it’s in an open AI tool, like ChatGPT, that information is effectively in the public domain. The SRA has been clear that using AI tools in this way will likely breach client confidentiality and as a result, legal professional privilege may be permanently waived and unable to be recovered. The SRA also stresses that both paid for and free-to-use AI tools pose risks in relation to client confidentiality, and the importance of firms understanding the safeguards that do (and do not) apply to the AI tools they are using.
The case law backing this up:
These aren’t hypothetical risks. Four recent cases show exactly how seriously the courts are taking concerns about the misuse of AI:
- Ayinde v Haringey LBC: fabricated case citations submitted to court led to a wasted costs application and referrals to professional regulators. The judgment is clear that reliance on AI is not a defence.
- BCP v A Mother: an unregistered barrister included AI hallucinations in submissions. No deliberate intent to mislead was found, but the court still treated this seriously.
- Cork v Smith: the court again reinforced that professionals are responsible for verifying AI-assisted work, hallucinations included.
- UK v Secretary of State for the Home Department: this is the case where the Tribunal directly addressed the confidentiality risk from using public AI tools. The Tribunal confirmed that ‘to put client letters and decision letters from the Home Office into an open source AI tool, such as ChatGPT, is to place this information on the internet in the public domain’.
The bottom line on accountability.
AI has no legal personality. However the work was prepared, you remain fully accountable for its accuracy, and that accountability extends to supervisors too. If a junior colleague submits AI-hallucinated content without proper review, the supervising solicitor can be found in breach of their own regulatory duties.
What firms should actually do about it?
- Verify everything, always. Every citation, every case reference, every “fact” an AI tool produces needs independent verification before it goes anywhere near a client or a court.
- Know where your data goes. Understand the contractual and technical safeguards of any AI tool before confidential information touches it, particularly where you are using free tools which may retain, store, or train on what you input.
- Supervise properly. Effective oversight of AI-assisted work isn’t optional, it’s an explicit regulatory expectation, with real consequences for supervisors who get it wrong. (Don’t forget that the SRA’s recently updated guidance on Effective Supervision includes a new section on supervision and AI).
- Build real governance, not just a policy document nobody reads. Your guidance to your firm needs to be proportionate, risk-based, and actually followed.
The takeaway.
AI is a genuinely useful tool for the profession, and it is likely to be used increasingly as new, more secure and efficient tools are developed. The SRA says as much, and is clear that AI (when used correctly) can be a valuable tool. But, and it’s a big but, the use of AI does not dilute professional responsibility. The judgement, the verification, and the accountability all still sit with the human lawyer.