Only what the model needs
We do not send the whole customer register to an AI service. Names and other details are removed where they do not change the answer.
STUDIO X AS develops systems, apps and AI solutions where security, privacy and operations are part of the delivery from day one. Data is stored in the EU/EEA, and we sign a data processing agreement when we process personal data on your behalf. AI only gets access to what it needs, what it does is logged, and a person approves anything that affects your customers. After launch we run the solution on a fixed agreement, so you have one supplier for both development and operations.
AI and the GDPR can work together. What matters is what is sent to the model and who is in control.
We do not send the whole customer register to an AI service. Names and other details are removed where they do not change the answer.
Which AI services are used, where the data is processed, and that it is not used to train the models, is written down before the feature goes live.
Answers that affect the customer, such as prices, status and email, are approved by a person or run in shadow mode until you see that they are right.
If the model is down, the feature falls back to manual use, not to an error message.
The solution is not locked to one AI vendor. A switch should take days, not months.
New AI features get limits per user, so an error or misuse does not lead to a large bill.
The AI models we use come from Anthropic (Claude) and OpenAI. Used directly, they may process data in the US. That happens under a data processing agreement with the EU Standard Contractual Clauses as the transfer basis, and the vendors do not use the data to train their models. If the data requires everything to stay in the EU, we use the same models with processing in the EU, for example Claude through a cloud provider with European data centres. We agree on what fits with you before we start.
We take responsibility for the solution: that it is built, secured and operated as the agreement describes, and that the AI features work as they are set up. An AI model can still give a wrong answer. That is why AI suggests and a person on your side approves anything that affects customers, finances or individuals.
A solution is production-ready when access control, logging, monitoring and an update plan are in place, not just when it works in a demo. That is the standard we deliver to.
We start with one well-defined process with clear rules and a measurable gain, for example interpreting incoming orders or documents. AI suggests, a person approves, and everything is logged. When the suggestions are right, more steps can run automatically. If the data is spread across several systems, we first bring it together in Central X.
Read more: how we build securely, our privacy policy and our terms.
Do you have security or privacy requirements in a procurement? Get in touch and we will go through them with you.

