On 17 August 2026, the SRA published its warning notice, Misuse of AI. It applies to all regulated firms and individuals using artificial intelligence in the provision of legal services and is also likely to be relevant to other authorised persons practising in SRA-regulated firms. The SRA states it will have regard to the notice when exercising its regulatory functions.
The SRA uses AI to mean technology capable of operating with a degree of autonomy and producing outputs not expressly determined or programmed in advance. Generative AI (“<span class="news-text_medium">genAI</span>”) refers to AI systems capable of creating new text, code, images, audio or video, including document summaries, drafting, research outputs and chatbot responses.
While recognising that many firms use AI safely and responsibly, the SRA warns that misuse can harm clients, disrupt court proceedings and undermine public confidence in the solicitors’ profession. An accompanying SRA news release reports 42 potential AI-misuse matters between July 2025 and July 2026, with ongoing investigations involving inaccurate citations, supervision and confidentiality.
The notice reflects the SRA’s outcomes-focused regulatory approach. Firms retain flexibility in how they deploy AI tools, but must ensure their arrangements meet the regulator’s professional standards.
The SRA identifies two principal areas of concern:
The SRA has received reports from senior members of the judiciary of potential Code of Conduct breaches, as well as self-reports from solicitors who relied on inaccurate or misleading AI-generated material. It also identifies instances in which confidential client information has been entered into public AI tools, creating risks to confidentiality and data-protection compliance.
AI has no separate legal personality. Solicitors and other regulated individuals remain accountable for their work and outputs, regardless of whether AI was used in its preparation. The notice highlights the duties to provide a competent service and effectively supervise work under paragraphs 3.2 and 3.5 of the Code of Conduct for Solicitors, RELs, RFLs and RSLs, as well as firms’ obligation to maintain effective governance, systems and controls under paragraph 2.1 of the Code of Conduct for Firms.
Solicitors must put forward representations that are properly arguable. They must ensure cited authorities are genuine and relevant, have verifiable citations and support the arguments advanced. The SRA recalls the observation in <span class="news-text_italic-underline">Brett v Solicitors Regulation Authority [2014] EWHC 2974 (Admin)</span> that misleading the court is “one of the most serious offences that an advocate or litigator can commit”.
The notice refers to <span class="news-text_italic-underline">R (Ayinde) v Haringey LBC [2025] EWHC 1383 (Admin)</span>, which concerned false AI-generated authorities filed in proceedings conducted under the court’s Hamid jurisdiction. The solicitor and barrister faced a wasted-costs application and regulatory referrals for conduct described as “improper” and “unreasonable”. The judgment stated a regulatory referral is likely to be appropriate where a lawyer places false citations before the court, whether through unverified use of AI or otherwise. Reliance on an AI output is not an answer: lawyers remain personally responsible for verifying every aspect of their submissions.
The SRA also cites <span class="news-text_italic-underline">BCP v A Mother [2026] EWFC 71 (B)</span>, in which an unregistered barrister was found to have misled the court through AI hallucinations, notwithstanding the absence of a deliberate intention to mislead, and <span class="news-text_italic-underline">Cork and another v Smith [2026] EWHC 1199 (Ch)</span>, which reaffirmed the responsibilities of authorised persons when errors arise from unchecked AI-generated content.
Putting false material before a court may also amount to contempt of court, potentially engaging paragraphs 2.5 and 2.6 of the Code of Conduct for Solicitors, RELs, RFLs and RSLs. It may interfere with the administration of justice and require courts to expend additional resources investigating and resolving the error.
Supervisors may themselves breach professional and regulatory obligations if false citations are submitted without adequate review or supervision. The notice highlights the obligations on solicitors managing or supervising others under paragraphs 3.5 and 3.6, the firms’ supervision duties in paragraphs 4.3 and 4.4 of the Code of Conduct for Firms and the compliance officer for legal practice’s duty under paragraph 9.1 to take reasonable steps to secure compliance.
It also notes that Rule 9.4 of the SRA Authorisation of Firms Rules requires regulated work to be supervised by at least one person with at least three years’ legal practice, whilst recognising not every supervisor must meet that requirement. Firms should nevertheless ensure each supervisor has appropriate experience.
The warning emphasises the risks of placing client information into public AI tools. In <span class="news-text_italic-underline">UK v Secretary of State for the Home Department [2026] UKUT 81 (IAC)</span>, the Upper Tribunal observed that entering client and Home Office decision letters into an open-source AI tool, such as ChatGPT, puts that information “on the internet in the public domain”.
The SRA states use of AI tools in this way is likely to breach client confidentiality and may permanently waive legal professional privilege. Both free and paid-for AI systems may present risk, depending on the supplier’s terms, system settings and technical architecture. Information entered into a tool may be stored, retained or used to improve the system, potentially resulting in loss of control over how confidential or client-sensitive information is processed and used.
Client information should therefore be entered into AI systems only where appropriate contractual, technical and organisational safeguards are in place. Firms and solicitors should satisfy themselves that data remains within a secure environment, cannot be accessed by unauthorised third parties, is not used to train AI models unless that use is expressly authorised and appropriate and is not retained longer than necessary.
The SRA stresses the importance of maintaining confidentiality under paragraph 6.3 of both Codes of Conduct, including where firms are deploying their own in-house AI tools. In-house solicitors must take particular care where their employer’s systems were developed for its commercial operations rather than for legal services. They should also consider whether the organisation’s interest in expanding its AI use could conflict with their personal professional duties.
The notice confirms AI may support legal practice, but does not dilute or transfer professional responsibility. Effective human oversight, informed professional judgment, proportionate risk management, robust supervision and secure information-handling arrangements are essential. The SRA warns that failing to have proper regard to the notice may result in disciplinary action.
<span class="news-text_medium">Source:</span> SRA, <a href="https://www.sra.org.uk/solicitors/guidance/misuse-ai/" target="_blank" class="news-text_link">Misuse of AI warning notice</a>, 17 August 2026.
