, Nils-Erik Jansson
AI incident response for legal teams
Within 72 hours of becoming aware, the department assesses whether it is a personal data breach, documents it and notifies the supervisory authority, unless the breach is unlikely to result in a risk to the data subjects. The rule is the GDPR. In Sweden the authority is IMY, the Swedish Authority for Privacy Protection, and the examples on this page are Swedish. It applies when someone pastes an HR matter into a personal account, uploads a customer list to the wrong project or enters a matter that is on the list of matters kept out. Here is the order: assessment, documentation, deletion, information, and what the vendor can and cannot do.
Updated by Nils-Erik Jansson
Hours zero to three
Stop and secure the facts
Who did what, when, in which tool, with which account, and which data about which people. Ask the member of staff not to delete anything until the facts are written down. Stop further use of the same kind until the assessment is complete. No question of blame at this stage.
Day one
The assessment
Three questions, on one page. The assessment is made by the person who owns the tool in the department, and the general counsel decides on notification. The data protection officer, if there is one, is informed the same day.
Days one to three
Notification, deletion, information
The notification to IMY is made by the controller. If information is missing when the deadline expires, the notification may be supplemented later. The deletion request is sent to the vendor in writing, stating the account, the time and the content. Record what the vendor confirms. The data subjects are informed in clear and plain language when the risk is high. For law firms, there is the further question of whether the client should be informed even when the GDPR does not require it.
72 hours from becoming aware
The deadline for notification
A late notification is worse than an incomplete one.
Always
The documentation
All breaches are documented, including those that are not notified: what happened, the assessment with reasons, the measures taken and what is changed to prevent it happening again.
The week after
Correct the rule as well as the person
Was the permitted tool too cumbersome, was the list of matters kept out unknown, were the project permissions too wide, or did the rule say no to everything so that people worked around it? Change what failed.
Background
The three typical cases
The first step is to secure the facts. Then you assess whether the event is a personal data breach, and if it is, IMY must be notified within 72 hours of becoming aware of it. There are three typical cases. The order below is a proposal.
Personal account. A member of staff has used their own account with an AI vendor for work tasks involving personal data. There is no data processing agreement, the consumer terms apply and there is no control over retention. This is the case the shadow AI guide deals with.
Wrong place in the right tool. The data was entered into the company's tool under the company's contract, but in a project or connector with the wrong permissions, so that colleagues who should not see the matter can see it.
Matter on the excluded list. The data was entered into the company's tool under the contract, but the matter is on the list of matters that may not be entered. One reason can be that the risk of a request from a public authority has not been accepted for that type of matter.
The three are assessed differently. The first is nearly always a breach to be assessed seriously. The second is a permissions issue that may be a breach. The third is a departure from the policy that is seldom a breach in the GDPR sense, because the processing takes place at a contracted processor. It must still be documented and corrected.
Hours zero to three: stop and secure the facts
Who did what, when, in which tool, with which account, and which data about which people. Ask the member of staff not to delete anything until the facts are written down, because a screenshot of the conversation is the only material there is for the assessment. Stop further use of the same kind until the assessment is complete. No question of blame at this stage. Whoever reports quickly has done the right thing.
Day one: the assessment
Three questions, based on IMY's description [1] and Articles 33 and 34 of the GDPR [2].
Is it a personal data breach? Has personal data been disclosed or made accessible to someone unauthorised, lost or altered? For a personal account, yes: the data is held by a vendor without a data processing agreement. For the company's tool under contract, usually no, unless the permissions have given access to unauthorised people.
Is the breach unlikely to result in a risk to the data subjects? Weigh the nature of the data (a name in a contract compared with health data in an HR matter), the scale, who the people are and what could happen to them. If the risk is unlikely, IMY does not need to be notified. Otherwise, notification must be made within 72 hours of becoming aware [1].
Is the risk high? Then the data subjects must be informed without undue delay [1]. The risk is lower for a single contract than for an HR matter with sensitive data.
| Question | What you weigh | Consequence |
|---|---|---|
| Personal data breach? | Whether data has been disclosed to someone unauthorised, lost or altered | No: document the departure. Yes: next question |
| Unlikely to result in a risk? | Nature of the data, scale, who the people are, what could happen to them | Yes: no notification, document. No: notify IMY within 72 hours |
| High risk? | The same factors | Yes: inform the data subjects without undue delay |
The assessment is written on one page with the three questions in the table, which should be ready before anything happens. The assessment is made by the person who owns the tool in the department, and the general counsel decides on notification. The data protection officer, if there is one, is informed the same day.
Days one to three: notification, deletion, information
The notification to IMY is made through IMY's online service by the controller, not by the processor [1]. If information is missing when the deadline expires, the notification may be supplemented later. A late notification is worse than an incomplete one.
The deletion request is sent to the vendor in writing, stating the account, the time and the content. What is actually deleted depends on the terms. On Anthropic's commercial plans, deleted chats are removed from storage within 30 days, while flagged content is kept for up to two years [6]. A personal account is governed by the consumer terms, and the undertaking not to train on the content does not apply there [5]. Record what the vendor confirms. Deletion limits the damage. It does not remove the incident from the assessment.
Information to the data subjects, when the risk is high, in plain language: what has happened, which data, what you have done, what they can do and whom they should contact.
For law firms, there is the further question of whether the client should be informed even when the GDPR does not require it. The Swedish Bar Association's guidance on generative AI takes the duty of confidentiality as its starting point [3], and disclosure of client information is a question of good professional conduct for advokater (members of the Swedish Bar Association) whatever the level of risk to the data subjects. That assessment is the firm's own.
Always: the documentation
All breaches are documented, including those that are not notified [1]: what happened, the assessment with reasons, the measures taken and what is changed to prevent it happening again. The documentation is what IMY asks for in a supervisory review, and it is the basis for the next step.
The week after: correct the rule as well as the person
After the incident, go through what in the procedure did not work. Was the permitted tool too cumbersome, was the list of matters kept out unknown, were the project permissions too wide, or did the rule say no to everything so that people worked around it? Change what failed. See the guide on an AI policy for the legal department and, if the problem is personal accounts, the guide on shadow AI.
A template for the assessment, a reporting route for staff and a procedure for who does what should make 72 hours enough.
How HOVR Legal works with this
The incident procedure, with the one page assessment, the reporting route and the allocation of roles, is part of the guidelines in the data protection package. The vendor's deletion procedure is documented in the review of the contract, see the guide on contracts with AI vendors. For Claude, it is part of the setup.
Frequently asked questions
- Is it a personal data breach every time someone has entered personal data into an AI tool without a contract?
- It is a security event that must be assessed as a possible breach. IMY describes a personal data breach as a security incident that leads to accidental or unlawful destruction, loss, alteration or unauthorised disclosure of, or access to, personal data [1]. Data given to a recipient without a data processing agreement has been disclosed to an unauthorised party. If the breach is unlikely to result in a risk to the data subjects, it need not be notified, but it must still be documented [1].
- Do the 72 hours run from the event or from when we found out?
- From when you became aware of the breach [1]. That is a reason to have a simple way for staff to report without being punished: the earlier the department knows, the more time there is for the assessment. The deadline is short enough that the assessment must be possible for whoever is available, using a template.
- Can the vendor delete what was entered?
- Partly, and it takes time. On Anthropic's commercial plans, deleted chats are removed from storage within 30 days, but content flagged by the safety systems is kept for up to two years and is not covered by the deletion [6]. A personal account is governed by consumer terms with different rules. Request deletion in writing, record what the vendor confirms, and do not assume that deletion undoes the incident for the purposes of the assessment.
Statements about Anthropic's terms and products were read on Anthropic's pages on 10 November 2026 and are checked at the start of every engagement.
Sources
- [1]Personuppgiftsincidenter, IMY, in Swedish; IMY's page on personal data breaches
- [2]Förordning (EU) 2016/679 (GDPR), artikel 33 och 34, EUR-Lex, in Swedish; the GDPR, Articles 33 and 34
- [3]Vägledning om användning av generativ AI i advokatverksamhet, Sveriges advokatsamfund, 12 juni 2025, in Swedish; Swedish Bar Association guidance on generative AI
- [4]Where are your servers located?, Privacy Center, Anthropic
- [5]Commercial Terms of Service, Anthropic
- [6]How long do you store my organization's data?, Privacy Center, Anthropic, read 2026-11-10
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.
- Data protection checklist for generative AI in legal work
A checklist for law firms and legal departments: five questions to answer before the first client data goes into an AI tool, and what counts as a pass.
- 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.
- 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.