Data protection law and AI: what lawyers must document
By the ChamberOne team 5 minute read
Data protection law differs between Sri Lanka, the UK, the US, Australia, Singapore and India, and it keeps changing. This article does not say what any of them requires. It describes the paperwork that a careful firm tends to keep when it uses AI tools, because regulators, clients and insurers usually ask about the same things: what do you hold, where does it go, who handles it, how long do you keep it and what happens if it goes wrong.
Key points
- Build records around questions every privacy regime raises, then take local advice on the specifics.
- Keep a register of AI tools, a vendor file, a retention rule and a breach plan that covers vendors.
- Log what is sent to which tool so you can trace an incident.
- Review and test the records against practice, not just against a template.
Start from the questions, not the statute
A compliance lead cannot memorise six regimes. A better approach is to build records that answer the questions every regime tends to raise, then take local advice on which of them the law requires you to keep and in what form. Check the guidance from your regulator or bar, and your data protection authority, on the rules that apply to you.
The records below also help with ordinary business needs: client questionnaires, insurer proposals and tender documents increasingly ask for them.
1. A register of AI tools and uses
List each AI tool the firm uses, the purpose, who approved it, which teams use it and what categories of information go into it. Include tools built into other software, such as a drafting assistant inside a word processor or an AI feature in a practice management system. Staff using personal accounts on public tools is a risk to find and address, so ask.
Review the register at set intervals and when a vendor changes its terms.
2. Records of processing
Many privacy regimes expect an organisation to know what personal information it processes and why. For AI use, note the types of data involved (client names, health details, financial records, children's data), the purpose, the legal basis or justification you rely on, who receives the data and where it is processed. Keep the record short and accurate, not long and stale.
Where processing is likely to carry higher risk, consider a written assessment of the risks and the steps taken to reduce them. Take advice on when your regime calls for one.
3. Vendor terms and due diligence
Keep a file for each AI vendor with the documents and answers that matter.
- The contract or terms of service in force on the date you adopted the tool, and later versions.
- Any data processing terms, including the vendor's role and its sub-processors.
- Where data is processed and stored, and whether it leaves your country.
- Whether your inputs are used to train or improve models, and how that is controlled.
- Retention and deletion terms, and how you can have data deleted.
- The vendor's security information and how it will tell you about an incident.
- Your due diligence notes: who asked what, and what was answered.
4. Retention and deletion
AI creates new records: prompts, outputs, logs and uploaded files. Decide how long each is kept, where, and who can see it. If a tool stores prompts, treat them as client information. If outputs are placed on the matter file, they follow the matter's retention rules.
Write down the rule and test it. Can you actually delete a client's uploaded document from the vendor's system? If not, say so in your notes and weigh that risk before using the tool for sensitive matters.
5. Notices and client information
Clients and other individuals may need to be told how their information is used, including that it may be processed by AI tools and by whom. Review your privacy notice and engagement terms against your register. Where a client has objected, or the matter is sensitive, record the decision not to use AI on it.
Keep a copy of each version of the notice and the dates it was in force.
6. Breach steps
A written incident plan should cover AI vendors as well as your own systems.
- 1
Who is the contact at each vendor, and how will they notify you of an incident?
- 2
Who in the firm receives that notice and who decides what to do?
- 3
How do you work out which clients' information was involved? A log of what was sent to which tool makes this possible.
- 4
Who decides whether to notify regulators, clients or others, with what advice and by when? Take advice on the deadlines that apply to you.
- 5
How will you record what happened and what you changed afterwards?
Keeping the paperwork honest
Records are only worth keeping if they match reality. Assign an owner, set a review date and test one record a quarter against what the firm actually does. If staff are using tools that are not on the register, the register is describing a firm that does not exist.
Do not copy a template and fill in generic words. A short record that says exactly what you do will serve you better in front of a regulator than a long one that reads like boilerplate.
- 1Assign an owner and a review date
- 2Test one record a quarter against what the firm does
- 3Find tools or uses missing from the register
- 4Correct the record
- 5Review again at the next date
- Then back to step 1.
A hypothetical example
Suppose a six-lawyer firm is asked by a corporate client to describe how it handles personal data in AI tools. The compliance lead opens the register and finds three approved tools, each with a vendor file, a retention rule and a note of where data is processed. The client's questionnaire is answered in an afternoon.
Compare the firm that cannot answer. It has to ask every fee earner what they use, discovers a free chatbot on two laptops and spends a week reconstructing what was pasted where. The paperwork did not make the first firm safer by itself, but it made the first firm able to say what it does, and to notice quickly when something changes.
- Opens the register and finds three approved tools
- Each has a vendor file, a retention rule and a note of where data is processed
- Answers the questionnaire in an afternoon
- Has to ask every fee earner what they use
- Discovers a free chatbot on two laptops
- Spends a week reconstructing what was pasted where
Proportion for a small firm
A two-partner practice does not need a forty-page programme. A single page for each of the tool register, the vendor file, the retention rule and the incident plan, kept up to date, is better than a thick binder nobody reads. Start with the tools that touch the most sensitive client information and widen from there.
Audit trail and AI controls in ChamberOne
- Every read and write is recorded automatically with who, when, IP address and before and after values. Users cannot switch it off.
- Every AI call is recorded in the audit log with the model and token counts.
- Audit events are retained for 7 years by default, on every plan, and can be filtered and exported.
What it does not do
- It does not tell you what data protection law requires of your firm or make your firm compliant.
- Tamper-evident hashing is on the roadmap and is not in the product today.
- With AI features on, the text needed for a request is sent to the AI provider for that feature; with them off, nothing is sent.
This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.
Read next
Corporate Transparency Act: who must still file beneficial ownership reports
FinCEN's August 2026 final rule made permanent the end of BOI reporting for US-formed companies; only certain foreign entities still file.
5 minute readClient confidentiality and AI tools: a checklist before you paste anything in
The questions to answer about retention, training use, access, hosting and sub-processors before client information goes into any AI tool.
5 minute readShadow AI in law firms: staff are already using it, so what next?
Why bans fail, how to find out what is being used, and how to offer an approved alternative with clear rules.
5 minute read