AI-generated GDPR access requests: what companies must do
Access requests under the GDPR are getting longer, more detailed and more legal in tone, and a chatbot often seems to have helped write them. Companies need to know whether they must answer such letters and how to keep to the deadlines.

What the supervisory authority figures show
The Hamburg data protection commissioner (HmbBfDI) received 4,220 complaints in 2025, up from 2,607 the year before, an increase of just over 60 percent. In the foreword to its 34th activity report, the authority names AI chatbots as one reason. They help people word their complaints and make them realise that a problem has something to do with data protection.
Hamburg is not alone. The commissioner in Lower Saxony counted about 70 percent more complaints in 2025 and also considers AI writing tools a possible cause. In Baden-Württemberg the number of complaints nearly doubled.
These figures describe complaints to authorities. The consultancy Dr. Datenschutz reports that it sees the same trend in requests sent directly to companies.
What AI-generated access requests look like
The rights of access (Art. 15), rectification (Art. 16), erasure (Art. 17) and objection (Art. 21) have existed in this form since 2018. Many people did not know exactly what they were entitled to, or avoided the effort of writing a letter. A short prompt now produces a finished request with article citations within seconds.
Typical features of such letters are:
- long question lists on data categories, recipients, storage periods and internal notes
- specific time periods and several demands in one letter
- references to statutes and judgments that do not always fit the case
- self-imposed deadlines shorter than the legal one
After the reply, the answer often goes back into the chatbot, which looks for gaps and drafts the follow-up question or complaint. The HmbBfDI points out that AI can lead people to assume too quickly that they are in the right. Art. 15 does not give a right to receive complete files, for example, even if a letter claims so.
Do you have to answer an AI-generated letter?
A request does not lose its legal effect because software drafted it, and a long or legalistic text is not abusive by itself. The content has to be examined and answered like any other request.
Accepting every legal claim in the letter without checking is equally wrong. You answer what the GDPR actually requires, and you can state politely where a demand goes beyond it. Companies commonly document that reasoning so they can show it to an authority later.
Late, incomplete or unclear answers risk a complaint to the supervisory authority. In one Hamburg case a retailer left access, erasure and direct-marketing objection requests unanswered beyond the one-month deadline for a long time, often until the authority stepped in. The HmbBfDI imposed a fine in the low six-figure range. Many late replies end differently, yet recurring process failures can have serious consequences.
Deadlines and review steps in practice
Under Art. 12(3) GDPR you must answer without undue delay and at the latest within one month. An extension by two further months is possible only for complex or numerous requests, and the reasons must be communicated within the first month. A deadline set by the requester does not replace the legal one.
A reliable process usually covers at least these points:
- central logging of all requests, whatever channel they arrive through
- recording the date of receipt and monitoring the deadline
- fixed responsibilities and escalation paths
- an appropriate identity check where there are justified doubts (Art. 12(6))
- involvement of all relevant systems and service providers
- a check of third-party rights before disclosure, with redaction where needed
Templates speed up the work but do not replace a look at the individual case. If you use AI tools internally to analyse requests or draft replies, review them under data protection law first. Personal data from a request does not belong in freely available AI services.
How company knowledge makes the response easier
Most of the effort lies in finding where personal data sits: mailboxes, file servers, project tools and chat. Searching them with knowledge management software shortens that step, provided the tool respects existing permissions. When each hit names its document, version and page, the person reviewing can verify it, as described on the page about answers with sources.
souveraen.ai connects sources such as SharePoint, Exchange and file servers read-only, with the existing permissions, through its integrations. It can be hosted in Germany or run on-premise. Agents prepare tasks, and anything that writes to a system waits for approval.
Whichever tool you pick, the review stays with a person. Check the GDPR basics on the page about GDPR-compliant AI before personal data enters any AI workflow, and agree who answers requests, who approves redactions and who watches the clock.
Sources
- Betroffenenanfragen nach DSGVO: KI erhöht den Aufwand, Dr. Datenschutz


