When automated scoring is allowed under Art. 22 GDPR
Companies that use bonitaet checks or scoring models need to know when Art. 22 GDPR applies and which information and objection rights follow. The same questions arise when AI systems rate customers or job applicants.

Why the question is on the table now
Creditworthiness scoring is routine in online retail, banking and rental decisions, and AI models are now entering the same processes. A recent article by Dr. Datenschutz from November 2026 summarises how far companies may go. It points to two rulings of the Court of Justice of the EU (CJEU) that shape current practice.
One question sits behind all of this: when does a score stop being a mere input and become a decision about a person? Anyone who buys scores from an agency, builds their own model or lets an AI tool rank applicants should be able to answer it for each process.
When does Art. 22 GDPR apply?
Art. 22 GDPR says a person should not be subject to a decision based solely on automated processing if it has legal effect or similarly significantly affects them. Not every score falls under this rule. The situation matters, as does the weight the score carries in the final decision.
In the Schufa ruling (C-634/21), the CJEU held that creating a score can itself count as such a decision. That applies when the score is decisive for a decision on a loan or contract and the person is significantly affected. A company that merely receives the score can therefore be in scope, even though an agency calculated it.
The significance threshold is assessed case by case. According to the source, restricting an online shop customer to safe payment methods is mostly not seen as reaching it, because the contract can still be concluded. A full refusal of a loan without a comprehensible explanation is different. The Berlin data protection authority fined Deutsche Kreditbank AG 300,000 euros because an algorithm rejected a credit card application without a clear explanation.
Which information and access duties apply?
Articles 12 and the following provisions of the GDPR require transparency toward the person concerned. Companies using profiling should be able to provide the following:
- Meaningful information about the logic involved
- The scope and intended effects of the processing
- The relevant factors and how they are weighted
- An individual explanation of the concrete decision
In the Dun & Bradstreet Austria case (C-203/22), the CJEU stressed that complexity does not release the controller from explaining. People must be able to follow the main steps and principles, so they understand which personal data were used and how they shaped the result. Saying only that an internal scoring model was used will usually fall short.
In practice, the department that uses the score, whether sales, finance or HR, is often the first contact for such requests. It needs access to the model documentation, and the data protection officer alone cannot supply that.
What changes when AI is used?
AI-based scoring raises the bar. An assessment should rest on correct facts and be explainable with a sound core statement. Pure machine learning black boxes whose logic cannot be explained sit uneasily with the GDPR transparency requirements.
The source recommends human review mechanisms, suitable and current data, and a way for affected people to contest the decision. Where AI is involved, it also advises close coordination with the data protection officer and a data protection impact assessment under Art. 35 GDPR. A human who only confirms the output without real scope to change it is unlikely to take the process outside Art. 22.
- List every process where a score or AI rating influences a decision about a person.
- Check how much weight the score carries and how severe the consequence is.
- Name a reviewer with real authority to deviate from the result.
- Prepare an explanation of factors and weighting that a layperson can follow.
- Set up a route for objections and document the outcome.
Using company knowledge for such reviews
Checks like these depend on finding model descriptions, vendor contracts, impact assessments and past decisions quickly. The documents often sit in several places, such as SharePoint, file servers and mailboxes. Teams that keep this material findable can answer information requests faster.
souveraen.ai is an AI platform for company knowledge from clouditiv GmbH in Auerbach, Germany. It answers questions from connected sources and cites document, version and page, so staff can verify each statement. It reads SharePoint, file servers and mailboxes with existing permissions. It runs either hosted in Germany or on-premise as Private Spark. Our page on the GDPR side of AI at work covers data protection, and answers with sources are explained separately.
Sources
- Automatisiertes Profiling in der Bonitätsauskunft, Dr. Datenschutz


