Verb
Back to blog
Safety2 min read

Is an in-app AI agent GDPR compliant? The questions to ask

The six questions to ask any vendor before an agent handles your users' personal data.

Quick answer

An in-app AI agent can be used in a GDPR-compliant way, but compliance depends on how the vendor handles your users' data, so ask. Check what it stores and for how long, whether one user's data can be deleted on request, which providers process it and where, whether conversations are used to train models, how your users are identified, and whether the vendor signs a data processing agreement. This is not legal advice; your own counsel decides.

Not legal advice: your own counsel decides what GDPR requires of you. What follows is the list of questions that decides it, and how Verb, which I build, answers each.

An AI agent inside your product handles personal data. Users type names, emails and account details, and an agent that acts reads data from your product as that user. Under GDPR you are usually the controller and the vendor running the agent is your processor. Whether the setup is compliant comes down to how that vendor handles the data.

1. What does it store, and for how long?

Ask for the list: conversations, user identifiers, action logs, analytics. Then ask how long each is kept. Verb keeps conversations until you ask for them to be deleted, so the record of what happened stays answerable, and you can export them as CSV or JSON at any time.

2. Can one user's data be deleted?

The right to erasure means you need to delete one person, not only a whole account. With Verb you send the user ID your backend puts in that user's token; their rows leave the live database at once, and the nightly backups that still hold them are gone within 14 days.

3. How are your users identified?

The less identifying data the vendor holds, the better. Verb receives a token your backend mints, verifies it, hashes it with a salt unique to your site and never stores it, so the same person using two products cannot be linked between them.

4. Which providers process it, and is it used for training?

An agent sends conversation content to a model provider to produce a reply. Ask which provider, where, and whether anything is used for training. Verb sends conversation content to an inference provider only to produce a reply, and it is not used to train anything.

5. Where does it run?

Ask for the hosting region and how one customer's data is kept apart from another's. Verb runs on its own servers in one region, with tenant isolation enforced by the database rather than by application code.

6. Will the vendor sign a data processing agreement?

Article 28 expects a written contract with your processor. Ask for it early, because it takes time on both sides. For Verb, a signed agreement is part of an Enterprise conversation: write to vishnu.ps@askverb.com.

The verdict

An in-app AI agent can fit a GDPR-compliant product. Get written answers to the six questions above before launch, keep the agent acting as the user rather than an admin, and record the vendor in your records of processing. Verb's security page lists every control in place today.

Frequently asked questions

Does an AI agent process personal data?

Almost always. What users type can include names, emails and account details, and an agent that acts reads data from your product as that user. Under GDPR you are usually the controller, and the vendor running the agent is a processor acting for you.

Are conversations with an AI agent used to train models?

It depends on the vendor and its model provider, so ask, and get the answer in writing. Verb sends conversation content to an inference provider only to produce a reply, and it is not used to train anything.

Can one user's data be deleted?

It should be, because GDPR gives people the right to erasure. With Verb you send the user ID your backend puts in that user's token, their rows leave the live database at once, and nightly backups that still hold them are gone within 14 days.

Vishnu

Vishnu, founder of AskVerb @itsvishnups

Building the AI agent that lives inside a SaaS product and does the work its users ask for. Writing down what I learn along the way, including the parts I got wrong first.

More articles

Give your SaaS its own AI agent

It calls your own API as the signed-in user, confirms before it changes anything, and logs every action. Start with a free 30-day pilot, then we price it on what it actually does for your users.