Skip to main content

GDPR-compliant AI development from the first prototype

AI projects in Europe often stall at the data protection review, after the prototype is already built. We design AI systems so that data minimisation, legal basis, retention and data subject rights are part of the architecture from the start.

ISO 9001 & ISO/IEC 27001Certified quality and information securitySince 2014, made in GermanyDiese Seite auf Deutsch

Privacy as an architecture decision

The GDPR does not prohibit using language models with business data, but it does require a purpose, a legal basis, minimal data, appropriate safeguards and the ability to fulfil access and deletion requests. These requirements are much easier to meet when they shape the design: which data enters a prompt, where embeddings are stored, how long logs are kept and which provider processes what.

Textwash Pro anonymises texts with local machine learning before they reach any AI system, so personal data can often be removed from the processing chain altogether. kalAIDoCreate, the jocapps platform for enterprises and the public sector, adds governance, audit logging and model choice. Both are built by jocapps GmbH in Freiburg, which is certified to ISO 9001 and ISO/IEC 27001.

What to watch out for

  • Check whether your model provider may use inputs for training and how long it retains prompts. Business and API terms often differ from consumer products, and the settings must match your processing agreement.
  • Transfers to providers outside the EU need a valid basis, such as an adequacy decision or standard contractual clauses. Document this per provider, including sub-processors.
  • A DPIA is often required for AI systems that process data at scale or evaluate people. Start it early, because its findings can change the design.
  • Pseudonymised data is still personal data under the GDPR. Only properly anonymised data falls outside its scope, so verify the anonymisation quality for your document types.
  • The EU AI Act adds separate obligations such as transparency and, for high-risk uses, risk management. GDPR compliance does not automatically cover them.

What privacy by design gives you

Less personal data to justify

Anonymising inputs with Textwash Pro before prompting or indexing reduces the scope of the data protection assessment. Data that never reaches the model does not need to be deleted from it.

Documentation your DPO can use

We supply data flow diagrams, processor lists and technical measures in a form that fits records of processing and a DPIA. Approval cycles become shorter and more predictable.

Control over providers and regions

kalAIDoCreate lets you choose models and hosting that match your requirements and switch later without rebuilding the application.

Deletion that actually works

Vector indexes, logs and caches are designed so that a deletion request can be traced and executed across all stores.

How we build GDPR-compliant AI systems

  1. Purpose and data map

    We document the purpose, data categories, sources and recipients together with your data protection officer.

  2. Architecture choices

    Anonymisation, hosting region, providers and retention periods are decided before development starts.

  3. Build with safeguards

    Access control, logging, prompt filtering and deletion routines are implemented and tested like any other feature.

  4. Assessment and operation

    We support the DPIA where required and set up reviews for model or provider changes.

Matching products

Textwash Pro

Local, automated text anonymization. Anonymize documents automatically with local machine learning before they reach any AI system. Privacy by design for regulated teams.

Discover Textwash Pro

kalAIDoCreate

Governed AI platform for enterprises and the public sector. An all-in-one AI platform for enterprises and public institutions with compliance features, governance and custom APIs.

Explore kalAIDoCreate

Frequently asked questions

Can we use US-based language models under the GDPR?

Often yes, with a processing agreement, a valid transfer mechanism and appropriate settings for data retention and training. Whether it fits your case depends on the data involved, which is why minimising and anonymising data first helps.

What does Textwash Pro do exactly?

It detects and replaces personal data such as names, addresses and identifiers in texts, using machine learning that runs locally. The anonymised text can then be passed to AI systems.

Do we need a DPIA for every AI project?

Not always. It is required when processing is likely to result in a high risk to individuals, which is common for AI systems working with personal data. Your data protection officer makes the final assessment.

Do you provide legal advice?

No. We provide the technical design, documentation and implementation. Legal assessment remains with your data protection officer or legal counsel.

Let's talk about your project.

15 minutes with an AI Code Engineer, free and without obligation. Or call our team around the clock.