This is a plain-English summary of published guidance, not legal advice.
What the SRA has published on AI
Three SRA publications set out its position, listed here from most recent to oldest.
- Warning notice: Misuse of AI (17 August 2026). The most recent and most direct. It names two problems: AI-invented case citations put before courts, and client information entered into public or poorly safeguarded AI tools. It says “reliance on an output of AI would not be a suitable defence”, and that misusing AI can waive privilege for good. The SRA says it will have regard to the notice when it uses its regulatory powers.
- Compliance tips for solicitors regarding the use of AI and technology (last updated 9 February 2026). Practical points: your COLP is responsible for compliance when new technology comes in, do due diligence on third-party platforms, and make it clear to clients when they are dealing with AI.
- Risk Outlook report: the use of AI in the legal market (20 November 2023). The background. It says not to trust an AI system to judge its own accuracy, to supervise both the systems and how staff use them, and that firms stay responsible for AI outputs. It also suggests telling clients when AI will be used on their case.
The SRA regulates outcomes rather than specific software. In May 2025 it authorised Garfield.Law, the first purely AI-based firm approved to provide regulated legal services in England and Wales, with solicitors remaining accountable for its work. The Law Society's Generative AI: the essentials is useful companion reading, though it comes from the representative body rather than the regulator.
The Code of Conduct duties AI touches
The warning notice maps AI onto the Code of Conduct for Solicitors and the Code of Conduct for Firms. These are the paragraphs that come up most.
- Confidentiality (6.3 in both Codes). Keep the affairs of current and former clients confidential. Pasting a client's documents into a personal AI account is the obvious way to breach it.
- Competence (Solicitors 3.2, 3.3 and 3.6; Firms 4.2 and 4.3). The service must be competent, and you and the people you manage must keep your knowledge up to date. That now includes knowing what your AI tools get wrong.
- Supervision (Solicitors 3.5; Firms 2.3 and 4.4). You stay accountable for work done through others and must supervise it. AI drafting done by a trainee is still the supervisor's work to check.
- Systems and controls (Firms 2.1 and 9.1). The firm needs effective systems and controls, and the COLP must take reasonable steps to make sure the firm complies.
- Client information (Solicitors 8.6). Give clients information they can understand, so they can make informed decisions about how their matter will be handled.
- The court (Solicitors 1.4, 2.4 and 2.7). Don't mislead the court, only make properly arguable submissions, and draw its attention to relevant cases. The notice cites R (Ayinde) v Haringey LBC [2025] EWHC 1383 (Admin), where fake citations led to wasted costs and referrals to the regulators.
UK GDPR and the ICO
If an AI tool processes personal data, UK GDPR applies. The ICO's Guidance on AI and data protection sets out how. The ICO says that guidance is under review after the Data (Use and Access) Act 2025, so check it for changes.
- Your firm is the controller. The AI provider is usually your processor, so you need a written contract.
- Assess the risk before you start. Using AI on client data will often call for a data protection impact assessment.
- Know where the provider stores and processes data, and whether it leaves the UK.
- Say in your privacy notice that you use AI tools and why.
- If personal data leaks through an AI tool, the usual breach rules apply, including reporting to the ICO within 72 hours where it's required.
Legal professional privilege
Privilege depends on confidentiality. The SRA's notice cites UK v Secretary of State for the Home Department [2026] UKUT 81 (IAC), where the Upper Tribunal warned that uploading client material to an open AI tool could put it in the public domain and waive privilege. The notice says privilege “may be permanently waived and unable to be recovered”.
In practice, privileged documents should only go into tools running inside your own systems, backed by contract terms that prohibit model training and limit data retention.
What this means for a firm using AI
- Write an AI policy and make your COLP its owner. Our free AI acceptable use policy template is a starting point.
- Approve the tools people may use, and ban personal AI accounts for client work.
- Read each provider's terms: training on your data, retention, where data is held, sub-processors.
- Do a data protection impact assessment and update your privacy notice.
- Check citations, facts and figures against primary sources before sending work out.
- Name who reviews AI-assisted work, and record it on the matter.
- Tell clients in your client care letter how you use AI and how you protect their data.
- Train staff before granting access, and keep a training record.
- Decide now who handles an AI incident and when you would tell the client, the ICO or the SRA.
To see which of these your firm already has in place, score it against our AI readiness checklist. It takes a few minutes and covers data, systems, people, policy and UK regulation.
How we set AI up within it
We configure AI for law firms to handle first drafts while fee earners review and approve work before it reaches a client or the other side. Client data stays in your Microsoft 365 tenant, or our UK-hosted environment, using business AI tiers that do not train on your data. Drafts link back to their source documents and are filed to the matter, and we prepare an AI policy for your COLP to adapt and own.
