AI and SRA compliance: a solicitor's guide
Using AI doesn't change your professional obligations — it just creates new ways to breach them if you're careless. Here's how the SRA framework applies.
- The SRA Standards and Regulations apply to AI-assisted work exactly as to any other work — competence, supervision, confidentiality, best interests.
- You cannot outsource professional responsibility to an AI system; the work and the liability remain yours.
- Verification of AI output against authoritative sources is a professional duty, confirmed in Ayinde.
- Confidentiality and data protection require care about what client information you put into which tools.
The SRA has not written a separate rulebook for AI, and arguably doesn't need to: the existing framework already covers it. The Law Society's guidance and the SRA's own materials make the same point repeatedly — your professional duties apply to AI-assisted work regardless of the technology used, whether the AI was used by you or by someone you supervise.
Competence
The duty to provide a competent service applies whether advice is handwritten or AI-assisted. In the AI context that has two edges: you must be competent enough to use the tool sensibly and to spot when its output is wrong, and you must not let the tool's fluency substitute for your own knowledge. Commentators have noted the SRA has yet to spell out exactly what competence requires here — which means the safe course is to assume the traditional standard applies in full.
Supervision
The court in Ayinde drew the analogy directly: relying on AI research is no different from relying on the work of a trainee solicitor or pupil. You supervise it the same way — you check it. Output from an AI tool is a draft to be reviewed, never a finished product to be trusted unread.
Verification — the core duty
This is the one that bites. The Divisional Court held that those who use AI for legal research have a professional duty to check the accuracy of that research against authoritative sources before using it in their work. Fabricated citations and invented authority are the characteristic AI failure, and passing them on — to a client or a court — is a breach. We unpack the case in Ayinde v Haringey explained and the practical method in how to verify AI output.
The Lawyer's Prompt — book + 15 Claude skills
Everything in this guide, built into tools you install once. Solicitor-trained skills that draft, review and verify to your house style, with the Ayinde discipline built in — and the book that explains the standard behind them.
See what's included — £99 →Confidentiality and data protection
Putting client information into an AI tool engages your confidentiality duty and UK GDPR. Before you paste identifiable client material into any tool, you need to understand how that tool handles data — whether inputs are retained or used for training. This is a judgement you make before the work starts; the tool can't make it for you.
Free tool: build your firm's AI policy
One practical step toward compliance is a clear internal AI policy. Build one tailored to your firm in a few minutes with the free AI policy generator — it covers the SRA Code, the Ayinde duty, confidentiality and supervision.
Build your AI policy free →How the skills help
The skills in The Lawyer's Prompt are built around this framework rather than bolted onto it. Every skill works to one consistent standard, never invents authority, and flags anything it cannot verify with [VERIFY] — so the verification step is structural, not something you have to remember. That makes AI use easier to supervise and easier to defend. It does not replace your judgement; it supports it.
Next: verifying AI output and AI for solicitors.
Primary sources
The guidance this page is based on, straight from the regulators: