, Nils-Erik Jansson
Data protection checklist for generative AI in legal work
Five questions need answers before the first matter or client data goes into an AI tool. The checklist takes them one at a time: what the question means, which document answers it, and what counts as a pass. It works for Claude, Legora, Copilot or any other tool. It is based on GDPR, with the Swedish Bar Association's guidance as the reference for law firms.
Updated by Nils-Erik Jansson
Answered: 0 of 5
Who is the processor, and under which agreement?
0 of 3What is the basis for the transfer to a third country, and has it been assessed?
0 of 4How long is data kept, and who can see it?
0 of 3What happens if an authority requests the data?
0 of 1Which matters are kept out of the tool?
0 of 3The guide with what each question means, where it is documented and the common mistake follows below as background.
Background
1. Who is the processor, and under which agreement?
| Point | Content |
|---|---|
| What the question means | The vendor of a cloud or AI tool processes personal data on your behalf and is a processor. That requires a written agreement with the content set out in Article 28 GDPR [1]. Without it there is nothing to assess, and the tool may not be used. |
| What is required | The agreement is signed by the company or firm, not privately by an employee. It states the purpose, the categories of data, the sub-processors and what happens when the agreement ends. You know which legal entity is your counterparty (for customers in the EEA, Anthropic's Irish company) and which plan the agreement covers. |
| Where it is documented | The vendor's processor agreement, for Anthropic the Data Processing Addendum that forms part of the commercial terms [4], plus your notes on what it covers and what it does not. |
| Risk | If an individual lawyer has bought the tool on Team or Pro with a company card, consumer terms apply, or a plan without the settings you need, and the processor agreement is not the company's. |
2. What is the basis for the transfer to a third country, and has it been assessed?
| Point | Content |
|---|---|
| What the question means | A US vendor means a transfer to the US, even if data is stored in the EU, because access from a third country is a transfer [1]. The transfer needs a transfer tool under Chapter V GDPR and an assessment that the tool holds up. |
| What is required | You know which transfer mechanism applies (for Anthropic, the standard contractual clauses, because the company was not among the active participants in the Data Privacy Framework list when I searched it). The assessment is done, dated, signed by the person responsible and read by the data protection officer. It ends in a list of matter types that are not entered and a date for review. You have a plan for what happens if the adequacy decision for the US falls [1]. |
| Where it is documented | The standard contractual clauses in the processor agreement, with completed annexes, and your six-step transfer impact assessment; see the guide on the assessment. |
| Risk | The vendor's own statement that it "complies with GDPR" is used as the assessment. That statement is input to the assessment. |
3. How long is data kept, and who can see it?
| Point | Content |
|---|---|
| What the question means | Retention, deletion and access, at the vendor and on your side, for chats, projects, uploaded files and whatever is created in the tool. |
| What is required | You know the default (for Claude: data is kept until you delete it, and deleted chats are removed within 30 days [7]), which settings your plan offers (custom periods only on Enterprise, minimum 30 days), what the settings do not reach (content flagged by the safety systems, public artifact links) and who on your side is Owner and can change them. Periods are set per matter type. Permissions in connected systems have been reviewed, so the tool sees no more than the user is allowed to see. |
| Where it is documented | The vendor's documentation on retention and settings, and your own settings list. For Claude: Anthropic's pages on retention and on custom retention periods [7] [6]. See the insight on retention in Claude Enterprise. |
| Risk | A project's retention period takes precedence over the chat's, and projects are unlimited by default. If you have only set a chat period, you have not controlled anything [6]. |
4. What happens if an authority requests the data?
| Point | Content |
|---|---|
| What the question means | If an authority in the vendor's country demands your data: will you be told, will the vendor object, and what do you do then? The Swedish Bar Association's guidance on external IT services refers to the US CLOUD Act. Its model clause requires the vendor to inform the firm, tell the court that the information is covered by legal professional secrecy, and oppose disclosure [2]. |
| What is required | One of three: the vendor has undertaken by contract to inform and object; you have asked for that, been refused and documented that the risk is accepted for certain matter types; or the matters where the risk cannot be accepted are on the list of matters kept out of the tool. All three are acceptable answers. Without an answer, the point is not done. |
| Where it is documented | The vendor's terms and transparency reports, your addendum or your documented request for one, and the section on residual risk in the risk assessment. |
| Risk | The question is skipped because it feels theoretical. It must still be answered for the matters where it is not. |
5. Which matters are kept out of the tool?
| Point | Content |
|---|---|
| What the question means | The sum of the first four: which matter types are not entered at all, however good the settings are. |
| What is required | The list exists, is short enough to remember, is known to everyone in the department and has an owner. For law firms: it reflects when client consent is required under the Swedish Bar Association's guidance on generative AI, that is, when risks remain that cannot be addressed by contract or security measures [3]. The list is reviewed each time the assessment is reviewed. |
| Where it is documented | A list by matter type in the AI policy, see the guide on the policy, with a reference to the step in the assessment that justifies each entry. |
| Risk | The list only says "sensitive matters", which is not a list. Write out the matter types. |
When all five have answers
Then the first matter can go in. The documents are the processor agreement with notes, the transfer assessment, any data protection impact assessment, the settings list, the policy with the list of matters kept out, and the training material. That is what the data protection package consists of, and what is included in the setup of Claude for Legal.
Frequently asked questions
- In what order should the questions be answered?
- In the order they appear. The processor agreement decides whether there is anything to assess. The transfer assessment decides which matters can go in. Retention and access decide the settings. The question on authority requests decides what remains as risk. The list of matters kept out is the sum of the first four.
- How long does it take to answer all five?
- The time depends on the vendor's documentation, the review with the data protection officer and the decisions: which matters are kept out, and whether the risk of an authority request is accepted or should be negotiated.
- Does the checklist also cover personal AI accounts that staff use?
- It shows why they cannot be approved. A personal account has no processor agreement, no transfer assessment and no retention settings, so questions one to three have no answer. That is why the policy requires the company's own agreement; see the guide on shadow AI in the legal department.
Statements about Anthropic's terms and products were read on Anthropic's pages on 9 November 2026 and are checked at the start of every engagement.
Sources
- [1]GDPR-vägledningen, Sveriges advokatsamfund, 10 september 2026, in Swedish; Swedish Bar Association GDPR guidance, sections 2.9, 2.13, 11 and 12
- [2]Uppdaterad vägledning om användningen av externa IT-tjänster i advokatverksamhet, Sveriges advokatsamfund, 15 april 2026, in Swedish; Swedish Bar Association guidance on external IT services in law practice
- [3]Vägledning om användning av generativ AI i advokatverksamhet, Sveriges advokatsamfund, 12 juni 2025, in Swedish; Swedish Bar Association guidance on generative AI in law practice
- [4]Data Processing Addendum, Anthropic
- [5]Where are your servers located?, Privacy Center, Anthropic
- [6]Configure custom data retention controls for Enterprise plans, Anthropic
- [7]How long do you store my organization's data?, Privacy Center, Anthropic, read 10 November 2026
Insights on the topic
Back to AI for legal
More under AI for legal
- Data protection package for AI tools in legal work
Fixed-scope engagement: transfer impact assessment, impact assessment where needed, draft addendum, policy and training. For the AI tool you use or consider.
- AI operating model for the legal department
Fixed-scope engagement for legal departments. You get an operating model for how AI is used, governed and followed up, whatever the tool.
- AI literacy training for lawyers: two hours
Two hours of training for your legal department or firm on your own tools and documents, documented as a measure under Article 4 of the AI Act.
- AI policy for the legal department: what it should contain
Outline of an AI policy for a legal department or law firm: permitted tools, what may be entered, review, responsibility, excluded matters and training.
- AI policy template for the legal department
An editable AI policy template for a legal department or law firm in nine sections, with comments on what to adapt and why.
- The EU AI Act for in-house counsel: what applies to users
What the EU AI Act requires of a legal department that uses generative AI: AI literacy since 2025, transparency since 2026 and when the high-risk rules apply.
- Transfer impact assessment for AI tools, step by step
The six steps of a transfer impact assessment for a US AI tool: map, choose a transfer tool, assess the country, supplement, document, review.
- DPIA for generative AI: when it is required and how
When a legal department or law firm must carry out a DPIA under Article 35 GDPR for an AI tool, using the Swedish regulator IMY's nine criteria.
- Personal data in prompts: legal basis and information
When a lawyer writes about an employee, a counterparty or a customer in an AI tool, personal data is processed.
- AI incident response for legal teams
What a legal department does within 72 hours when a member of staff has entered data into an AI tool that should not have gone in.
- Anonymise before prompting: what works
How a lawyer removes personal data before text goes into an AI tool, and the difference between anonymisation and pseudonymisation.
- Swedish MBL and AI tools: must you negotiate first?
Swedish law: when an AI tool triggers the duty to negotiate under section 11 MBL, when informing under section 19 is enough, and with no collective agreement.
- Shadow AI in the legal department: one week of work
Staff use ChatGPT and Claude on private accounts for work. The legal department's week: map it, stop the worst, offer an approved tool, write the rule.
- Legal privilege and AI: the Swedish Bar guidance
What the Swedish Bar Association's guidance on generative AI requires for confidentiality: tools that do not train on client data, and terms read in full.
- Contracts with AI vendors: asking beyond standard terms
What a legal department or law firm asks of an AI vendor beyond the standard terms: authority requests, sub-processors, deletion, anonymisation, exit.
- On-premise or cloud AI for legal work
When a legal department or law firm should run AI locally and when the cloud is right: what local operation solves, what it costs, what it leaves open.
- Reviewing an AI answer in ten minutes: five steps
How a lawyer reviews an answer from Claude, Legora or ChatGPT before using it: sources, case law and statutes, figures, what is missing, and tone.
- AI literacy training for the legal department
What a two-hour training for lawyers using generative AI covers: what the tool does and does not do, your rules, practice on your own documents.
- AI guidance for lawyers in the EU and the UK
What the CCBE, the SRA, the Law Society, the Bar Council and the Swedish Bar Association require of lawyers using generative AI, and where they differ.
- The general counsel's AI report to the board: two pages
What the board needs to know about the company's AI use once a year, and what the general counsel can answer for.