Technology and AI

What the SRA expects from firms using AI

How existing SRA rules apply to AI: accountability, competence, supervision, confidentiality and client information, with an evidence file checklist.

Ava research teamPublished 6 min read

On this page
  1. No AI rulebook, but plenty of rules
  2. Accountability stays with the firm
  3. How the Code of Conduct applies
  4. What the first AI-driven firm's approval shows
  5. What clients expect
  6. Data protection sits alongside the Code
  7. Questions the COLP should be able to answer
  8. Building an evidence file
  9. Example: answering a regulator's question
  10. Common gaps
  11. Frequently asked questions
  12. References

Key takeaways

  • There is no separate AI rulebook for solicitors; the SRA Code of Conduct and existing duties apply to AI-assisted work.
  • Firms remain accountable for AI outputs, including errors made by a third-party chatbot speaking to clients.
  • Supervision, competence and confidentiality are the duties AI most often tests.
  • Keep an evidence file: approved tools, risk assessments, training, checking routines and incident logs.

No AI rulebook, but plenty of rules

Firms looking for a single SRA rule on artificial intelligence will not find one. The UK has taken a pro-innovation, light-touch and cross-sectoral approach to AI regulation, leaving existing regulators to apply their existing rules, as the Law Society's summary of the government's position explains[1]. For solicitors, that means the SRA Code of Conduct, the Accounts Rules, the Transparency Rules and the general law on data protection and confidentiality all apply to AI-assisted work exactly as they apply to any other work.

That is good news in one sense: the duties are familiar. The challenge is applying them to tools that behave differently from people and from traditional software. A trainee who is unsure says so; a generative tool rarely does. A precedent stays the same until someone edits it; an AI output can change each time it is produced.

Accountability stays with the firm

The SRA's Risk Outlook report on AI in the legal market is clear that firms remain responsible and accountable for outputs from AI they use, including errors in advice given by a third-party chatbot[2]. Buying a tool from a reputable supplier does not move responsibility to the supplier. If a chatbot on your website tells an enquirer something wrong, or an AI drafting tool introduces an error into a letter, the firm answers for it.

The same report is positive about what AI can do. It notes that legal chatbots can respond quickly to enquiries from potential clients, potentially offering a 24/7 service, gathering information and triaging cases for urgency[2]. The regulator is not asking firms to avoid AI. It is asking them to stay in control of it.

How the Code of Conduct applies

Selected paragraphs of the SRA Code of Conduct for Solicitors, in effect from 11 April 2025, applied to AI.
Code paragraphWhat it requiresWhat it means for AITypical control
1.4Do not mislead clients, the court or othersAI outputs presented as checked work must have been checked; automated systems should not pass themselves off as peopleVerification step; clear labelling of automated services
3.2Competent and timely serviceErrors from AI are still errors in your serviceReview before anything leaves the firm
3.3Maintain your own competenceUsers need to understand what tools can and cannot doTraining before access is granted
3.5Remain accountable for, and effectively supervise, work done through othersSupervision extends to work produced with AISampling, file notes and supervisor sign-off
3.6Ensure the people you manage are competent and up to dateManagers must make sure staff are trained on approved toolsTraining records and refreshers
6.3Keep client affairs confidentialClient data must not go into tools that do not protect itApproved tools list; contracts; settings
8.6Give clients information in a way they can understandClients may need to know how AI affects their matterClient care wording and a clear answer when asked

Supervision deserves the most attention. Under the Code, where you supervise or manage others providing legal services, you remain accountable for the work carried out through them and must effectively supervise work being done for clients[3]. In an AI context, effective supervision means knowing which tools your team uses, for which tasks, and how outputs are checked. A supervisor who cannot say whether a letter was drafted with AI cannot judge whether the checking was adequate.

Confidentiality is the second pressure point. The Code requires you to keep the affairs of current and former clients confidential unless disclosure is required or permitted by law or the client consents[3]. Pasting client details into a tool whose terms allow the provider to retain or reuse inputs is hard to reconcile with that duty.

What the first AI-driven firm's approval shows

On 6 May 2025 the SRA authorised Garfield.Law Ltd, the first purely AI-based firm authorised to provide regulated legal services in England and Wales[4]. The conditions are instructive for any firm using AI with clients. The approval required that the system only takes a step where the client has approved it, with named regulated solicitors accountable for all outputs[4].

  • Client approval before action: automated tools should not take steps on a client's behalf without the client agreeing.
  • Named accountability: a specific solicitor is responsible for what the system produces.
  • Scope limits: the approval covered a defined service rather than open-ended legal work.
  • Oversight by design: controls were built into the process, not added after problems appeared.

What clients expect

Regulatory duties line up closely with what the public says it wants. The Legal Services Board's 2026 public panel research found strong expectations of AI legal tools[5].

What legal service users expect of AI legal toolsLegal Services Board public panel, 2026; share of panel members
What legal service users expect of AI legal toolsBar chart. Expect a complaints process: 89%; Expect tools kept up to date with the law: 87%; Never act automatically for a vulnerable person in high-stakes cases: 81%; Rate accuracy as most important: 72%; Acting without knowledge or consent is most urgent concern: 56%.Expect a complaints processExpect a complaints process: 89%89%Expect tools kept up to date with the lawExpect tools kept up to date with the law: 87%87%Never act automatically for a vulnerable person in high-stakes casesNever act automatically for a vulnerable person in high-stakes cases: 81%81%Rate accuracy as most importantRate accuracy as most important: 72%72%Acting without knowledge or consent is most urgent concernActing without knowledge or consent is most urgent concern: 56%56%What legal service users expect of AI legal toolsBar chart. Expect a complaints process: 89%; Expect tools kept up to date with the law: 87%; Never act automatically for a vulnerable person in high-stakes cases: 81%; Rate accuracy as most important: 72%; Acting without knowledge or consent is most urgent concern: 56%.Expect a complaints processExpect a complaints process: 89%89%Expect tools kept up to date with thelawExpect tools kept up to date with the law: 87%87%Never act automatically for avulnerable person in high-stakes casesNever act automatically for a vulnerable person in high-stakes cases: 81%81%Rate accuracy as most importantRate accuracy as most important: 72%72%Acting without knowledge or consent ismost urgent concernActing without knowledge or consent is most urgent concern: 56%56%

Source: Legal Services Board, 2026[5]

Show the data
ItemValue
Expect a complaints process89%
Expect tools kept up to date with the law87%
Never act automatically for a vulnerable person in high-stakes cases81%
Rate accuracy as most important72%
Acting without knowledge or consent is most urgent concern56%

In that research, 89% of legal service users expected an AI legal tool to have a complaints process, 87% expected tools to be kept up to date with important changes in the law, and 81% agreed AI tools should never act automatically for a vulnerable person in high-stakes situations[5]. AI acting without the user's knowledge or consent was the single most urgent concern, named by 56%[5]. Firms that design around these expectations will usually find they have also met the Code.

Data protection sits alongside the Code

The SRA's rules are not the only ones in play. Most AI tools process personal data, so UK GDPR applies alongside the Code: the firm needs a lawful basis, a processor contract with the supplier, a privacy notice that tells people what happens to their data, and a data protection impact assessment where processing is likely to result in a high risk. Automated decisions that significantly affect people carry extra safeguards. In practice, the same evidence file can hold both the regulatory and the data protection records, and the COLP and the data protection lead should review new tools together rather than separately.

Questions the COLP should be able to answer

  1. Which AI tools are approved, and for which tasks?
  2. Which tools touch client data, and what do their contracts say about storage, retention and reuse?
  3. How is AI-assisted work checked, and how do supervisors know it was?
  4. What do clients and enquirers see or hear about our use of AI?
  5. What errors or near misses have we recorded, and what changed as a result?

If any of these takes more than a few minutes to answer, the gap is usually in record keeping rather than in practice. Closing it is mostly administrative work, and it is far easier to do before anyone asks.

Building an evidence file

If the SRA, an insurer or a client asks how the firm controls its use of AI, the answer should be a folder, not a conversation. Build it as you go.

  1. Approved tools register

    List each tool, its supplier, what it may be used for, what data may go into it and who approved it.

  2. Risk assessments

    Record a short assessment for each tool, including a data protection impact assessment where processing is likely to be high risk.

  3. Policy and training

    Keep the current AI policy, the training given and who completed it, with dates.

  4. Checking routines

    Describe how outputs are checked for each use, and keep samples of supervisor reviews.

  5. Client information

    Keep the wording used in client care letters, website notices and call scripts about AI.

  6. Incidents and changes

    Log errors, near misses and supplier changes, and what the firm did about them.

Example: answering a regulator's question

Example (fictional): following a complaint, a firm is asked how an incorrect statement came to appear in a letter to a client. The COLP opens the evidence file. The register shows the drafting tool was approved for first drafts only; the training log shows the fee earner completed training; the matter note records that the draft was reviewed, but by the fee earner alone rather than the supervisor required for that matter type. The firm can show what its controls were, where they failed and what it changed: supervisor sign-off is now enforced by the case management workflow for that matter type.

Common gaps

  • No list of approved tools, so staff use whatever is convenient, including personal accounts.
  • Supervisors unaware which matters involved AI-assisted work.
  • Website chatbots or call handling launched by marketing without compliance review.
  • Client care letters that say nothing about AI even where it materially shapes the service.
  • No log of errors, so the same problem recurs.

For the practical steps behind these points, see writing a firm AI policy, AI hallucinations and fake citations and generative AI in law firms. Platform settings that support confidentiality are covered in Microsoft 365 for law firms.

Frequently asked questions

Has the SRA issued specific rules on AI?

There is no separate AI rulebook. The SRA applies its existing Code of Conduct and other rules to AI-assisted work, and has published research and risk material on AI in the legal market.

Who is responsible if an AI tool gives a client wrong information?

The firm. The SRA has said firms remain responsible and accountable for outputs from AI they use, including errors in advice given by a third-party chatbot.

Do supervisors need to know when AI was used on a matter?

In practice, yes. Effective supervision under the Code requires knowing how work was produced so that the checking can be judged adequate.

What should a firm keep to show it uses AI responsibly?

An approved tools register, risk assessments, the AI policy, training records, checking routines, client information wording and an incident log.

References

  1. The Law Society (2026). Generative AI: the essentials (opens in a new tab) ↑

  2. Solicitors Regulation Authority (2023). Risk Outlook report: The use of artificial intelligence in the legal market (opens in a new tab) ↑

  3. Solicitors Regulation Authority (2025). SRA Code of Conduct for Solicitors, RELs, RFLs and RSLs (opens in a new tab) ↑

  4. Solicitors Regulation Authority (2025). SRA approves first AI-driven law firm (opens in a new tab) ↑

  5. Legal Services Board (2026). AI in Legal Services: Research Report (LSB Public Panel) (opens in a new tab) ↑

Book a demo

See Ava handle your calls.

Book a short demo. We’ll walk through your calls and workflow, and agree what Ava should take care of.

Book a demo