AI governance and adoption in UK legal services
Legal work is text-heavy, which makes it an obvious target for language models — and unusually exposed when those models are wrong. Fabricated citations, confidentiality breaches and unsupervised advice are the failure modes that reach regulators. This hub covers where AI helps legal teams and what supervision looks like in practice.
Who regulates this: SRA, Bar Standards Board, ICO, CILEX Regulation
Last reviewed: 18 August 2026
Where the risk sits
Fabricated cases and citations
Language models invent plausible case names, citations and quotations. Courts have criticised submissions containing non-existent authorities, and the responsibility sits with the professional, not the tool.
Confidentiality and legal professional privilege
Pasting client material into a consumer AI tool can breach confidentiality obligations and may raise questions about the handling of privileged material. The applicable duty depends on the firm's terms of engagement and the specific tool's data handling.
Supervision and competence
Regulated work produced with AI assistance still needs an accountable, competent human reviewer where the firm's supervision arrangements require it. Unsupervised output delivered to a client without that review is a conduct risk.
Client disclosure and billing
Clients increasingly ask whether AI was used and how it affects fees. There is no general rule mandating disclosure in every case, but telling clients how AI is used and how it affects fees is widely regarded as best practice and may be required by a specific engagement or costs rule.
Controls that make a rollout defensible
- Written AI use policy naming approved tools and prohibited data types
- Mandatory citation verification against a primary source before any authority is relied on
- Enterprise agreements that exclude training on your data and set retention limits
- Clear internal rule on who signs off AI-assisted work product before it reaches a client
- Client-facing statement covering AI use where your engagement terms or a specific rule calls for it
This page is guidance, not legal, clinical, financial or other professional advice. It is general information about UK regulatory context and does not account for your specific circumstances. Take professional advice before acting. See our editorial policy.
Sources
Frequently asked questions
Are UK solicitors allowed to use AI?
Yes. There is no prohibition. Existing duties on competence, confidentiality and supervision apply unchanged, so the firm must be able to show how those duties are met when AI is involved.
What is the biggest practical risk?
Unverified citations and quotations. A single fabricated authority in a filed document can cause professional and reputational damage that no efficiency gain offsets.