Using ChatGPT, Gemini or Claude on Conveyancing Files: What the SRA and the Courts Say

The Solicitors Regulation Authority (SRA) says client information should go into an artificial intelligence (AI) tool only where contractual, technical and organisational safeguards are in place, and that privilege may be permanently waived. The Upper Tribunal has said that putting confidential material into an open-source tool such as ChatGPT breaches confidentiality and waives privilege. The SRA says paid plans may carry the same risks.

I am Chief Operating Officer of Lex Sentinel, which builds AI tools for conveyancers. This article is general information for firms regulated by the SRA in England and Wales and is not legal advice. It draws on the sources listed below, which I checked on 25 October 2026. Regulator pages and provider terms change, so check the current versions, and I have not tested any named product. Statements marked "in my view" or "my suggestion" are mine, and the rest come from the sources.

Can a conveyancer put client information into a public AI chatbot?

A conveyancer should enter client information into an AI tool only where the firm has confirmed appropriate contractual, technical and organisational safeguards, and the SRA says both free and paid tools may not provide them. The SRA's warning notice on the misuse of AI, published on 17 August 2026, says client data should be entered into an AI tool only where appropriate contractual, technical and organisational safeguards are in place. It says input may be stored, retained or used to improve the tool. It sets out what the firm should be satisfied about: that the data stays in a secure environment, that it is not accessed by unauthorised third parties, that it is not used to train AI models except where explicitly authorised and appropriate, and that it is not retained longer than necessary.

The notice uses the terms public AI tools and free-to-use tools. It names ChatGPT once, as an example of an open source AI tool, and does not name Gemini or Claude. It says that both paid for and free-to-use tools may not provide the contractual and technical safeguards, and that the risk depends on the provider's terms, settings and technical architecture, so a firm has to check them for the plan it actually uses.

What happens to privilege and confidentiality?

Putting confidential client material into a public AI tool can breach confidentiality and waive privilege, and the SRA says a waiver of privilege may be permanent and unrecoverable. The Upper Tribunal went further in UK and R (Munir) v Secretary of State for the Home Department ([2026] UKUT 81 (IAC)), promulgated on 17 November 2025. The case arose in the immigration jurisdiction. At paragraph 21 the Tribunal said that putting confidential material into an open-source AI tool, such as ChatGPT, places it on the internet in the public domain, breaches client confidentiality and waives legal privilege, and that a firm which does this should consult the Information Commissioner's Office (ICO).

At the same paragraph the Tribunal said closed source AI tools that do not place information in the public domain, such as Microsoft Copilot, are available for tasks such as summarising without these risks. It did not set out a test for what counts as a safe tool, as DAC Beachcroft also noted on 18 March 2026.

The duty underneath is paragraph 6.3 of the SRA's Code of Conduct for Firms, which requires a firm to keep the affairs of current and former clients confidential unless disclosure is required or permitted by law or the client consents. Paragraphs 2.1 and 2.3 of the same Code require effective governance, systems and controls, and say the firm remains accountable where work is carried out through others.

What does data protection law add?

Data protection law adds duties about who counts as a controller, the lawful basis for each purpose of processing, and restricted transfers of personal data outside the UK. The ICO's innovation advice answers say an AI provider is likely to be a processor where it acts on the organisation's instructions, and a separate controller where it trains its own model for its own purposes. They also say an organisation must identify a lawful basis for each different purpose. The page records updates to December 2025 and is marked under review because of the Data (Use and Access) Act 2025.

On transfers, the ICO's guide to international transfers, updated on 15 January 2026, says the rules apply to all restricted transfers, even small or infrequent ones, and sets out three routes: adequacy regulations, appropriate safeguards and exceptions. The guide does not mention the Data (Use and Access) Act 2025, so check whether the Act changes the transfer test before relying on it.

The ICO's page on when to carry out a data protection impact assessment, dated 13 October 2025, lists AI as innovative technology and says an assessment is needed where processing is likely to result in a high risk. The ICO marks several of these pages as under review because of the 2025 Act, but its page on the Act says all the provisions affecting data protection law are now in force.

What does the Law Society say?

The Law Society's generative AI guidance, updated in September 2025, advises that you generally should not feed confidential information into generative AI tools, and says firms should check where data is processed, stored and accessed. Its article on conveyancing and AI, which carries no date, suggests asking vendors whether they will use your data to train their product and recording when AI was used on each matter.

What happens when an AI tool gets the answer wrong?

A chatbot can produce confident output that is wrong, and the firm remains responsible for it. Our guide to AI hallucinations in legal work covers the court decisions and the SRA's position. In my view, the conveyancing risk is a tool that invents a lender requirement, misreads a lease clause or misstates a search result, and the safeguard is the same: a named person checks the output against the source document.

What should a firm do?

A firm should decide which tools staff may use and for what, and then enforce it. The sources I have read do not prescribe a policy, so these steps are my own suggestions.

1. List the tools staff are allowed to use for client work and state in writing that others are not.
2. For each approved tool, check the provider's terms on retention, training on your data, access by third parties and where the data is processed.
3. Complete a data protection impact assessment before any tool processes client personal data in a way likely to be high risk.
4. Use synthetic or redacted data for experiments, and never real client documents in a tool the firm has not approved.
5. Require a named person to verify output against the source document before it is relied on.
6. Train staff and record the training, and record on the file when AI was used.

About the author

Appan Pathmanathan is a solicitor who has worked in UK legal practice since 2001, including litigation involving money laundering and fraud, and conveyancing, and has been managing partner of a law firm. He delivers anti-money laundering and compliance training. Profile: LinkedIn.

Sources

Checked on 25 October 2026.

1. SRA, Misuse of AI warning notice (17 August 2026)
2. SRA, Code of Conduct for Firms
3. [UK and R (Munir) v Secretary of State for the Home Department [2026] UKUT 81 (IAC)](https://caselaw.nationalarchives.gov.uk/ukut/iac/2026/81)
4. DAC Beachcroft, Using AI tools and legal privilege: a warning from the Upper Tribunal (18 March 2026)
5. ICO, Innovation advice: previously asked questions
6. ICO, A brief guide to international transfers (15 January 2026)
7. ICO, When do we need to do a DPIA? (13 October 2025)
8. Law Society, Generative AI: the essentials (September 2025)
9. Law Society, Conveyancing and AI: finding the right approach for your firm

Frequently Asked Questions

Can I put client documents into ChatGPT?

Only if the firm has confirmed appropriate contractual, technical and organisational safeguards. The SRA's warning notice names ChatGPT as an example of an open source tool and says both free and paid tools may not provide them, so check the provider's terms, settings and technical architecture for the plan you use.

Does using a public AI tool waive privilege?

It can. The SRA says privilege may be permanently waived if confidentiality is breached, and the Upper Tribunal in Munir said putting confidential material into an open-source AI tool such as ChatGPT breaches confidentiality and waives privilege.

Are paid AI plans safe for client data?

It depends on the provider's terms for that plan. The SRA's notice does not say paid tools are safe, and asks firms to check retention, unauthorised third-party access and use of data to train models.

Do I need a data protection impact assessment for an AI tool?

If the processing is likely to result in a high risk, yes. The ICO lists AI as innovative technology, and recommends doing one if in any doubt.

What should a firm's AI policy cover?

The approved tools and permitted uses, the checks on each provider's terms, a data protection impact assessment, rules on real client data, who verifies output, and training. These are my suggestions, as no regulator prescribes a policy.