GDPR and AI agents: what a small business has to think about
GDPR applies to AI agents as soon as they read or change personal data. For most business agents that is the case from the first task. A small business does not need a new set of rules for this. It needs to know four things about each agent: what data it uses, why, where the data goes and who is responsible.
This article is a general orientation and not legal advice. For your own case, ask a data protection adviser.
The terms in plain words
- GDPR is the EU's data protection law. It sets rules for how organisations handle information about people.
- Personal data is any information about a person who can be identified: a name, an email address, an order history.
- An AI agent is a program that carries out tasks for you with the help of AI, such as sorting customer emails or drafting replies.
Why agents deserve a second look
Think of an agent as a very fast new assistant with a key to the filing cabinet. Three things change.
- It reads more. An agent can read more customer files in a minute than a person opens in a week.
- Another company is involved. The AI behind the agent is often run by a different company from the one that provides your software. Personal data may pass to that company.
- It acts. An agent can change entries or prepare decisions about people.
Five things to think through
- Purpose. You collected the data for a reason. Check that the agent's task fits that reason. Answering a customer's question from their order history fits. Using old job applications for something new does not.
- Only what is needed. Give the agent access to the data its task requires and no more. An agent that drafts project updates has no need for salaries.
- Where and by whom. Find out which companies handle the data when the agent works, and in which region. You will normally need a data processing agreement with your software supplier. That is a contract on how the supplier may handle your data.
- People's rights. A person can ask what you hold about them and ask for it to be corrected or deleted. That includes notes and summaries an agent wrote.
- Decisions about people. Where a result affects a person seriously, for example in hiring, a human should decide and be able to explain the decision.
A short record for each agent
The simplest way to stay on top of this is a short record per agent. It answers most questions a customer, an employee or an authority could ask.
| Question | Example entry |
|---|---|
| What does the agent do? | Sorts customer requests and drafts replies |
| Whose data does it read? | Customers who wrote to us |
| Which details? | Name, email, message, order history |
| What is blocked? | Payment details |
| Where does the data go? | Software supplier and AI, both in the EU |
| Who approves its work? | The colleague on duty |
| Who is responsible? | Head of customer service |
Data that needs extra care
Some kinds of data are more sensitive under GDPR, for example health data. Employee data also deserves caution. If an agent would touch any of these, get advice first and keep its access as narrow as you can.
What about public chat tools?
Pasting customer data into a public AI chat can be a problem. You may have no agreement with that provider. Using AI inside software you have a contract with is the cleaner route.
How Worknano handles this
Worknano stores data in the EU (Frankfurt) by default. It keeps each customer's data fully separate and signs a Data Processing Agreement on request. With regional hosting, workspace data and AI processing stay in the chosen region. You decide how long data is kept, can export everything and can request deletion at any time. Agents act within the access rights of the person asking. Rules for single fields keep sensitive data such as salaries out of view.
- Is Worknano GDPR compliant?
- Where is our data stored, and is it secure?
- Permissions: roles, access and who can see what
Create a free workspace and see how permissions limit what an agent can read.