News
DPA warns debt collection agencies: AI requires human oversight
A sample of 29 debt collection agencies leads to recommendations on human intervention, personal data and transparency.
The Dutch Data Protection Authority (AP) published recommendations on 29 October 2026 for the use of algorithms and AI by debt collection agencies. The trigger was a sample of 29 agencies. According to the supervisory authority, AI is being used increasingly frequently, whilst the application can also create risks of incorrect decisions, discrimination and unequal treatment. The key message: human staff must remain involved in decisions with major consequences for debtors. [1]
From call transcript to probability calculation
The AP distinguishes between simple applications, such as chatbots and conversation summaries, and systems that go further in substance. Debt collection agencies use algorithms, for example, to estimate how likely it is that they will successfully collect a debt. The source mentions credit score, location and payment history as variables. Staff can then use such a calculation to determine which debtors they contact first. [1]
A prediction is therefore not merely an administrative summary: it can influence the order and approach of contact. This makes the quality of the input and the human assessment relevant. This is editorial interpretation of the application described; the AP publication does not prove that every agency uses the same method or makes the same mistakes.
Enriching files requires attention to data use
Another application described is file enrichment. AI or algorithm systems then search for additional data in registers or on the internet, for example contact details, bank details, company information or social media activity. This information can subsequently end up in a file and be used for communication or estimating the collection probability. [1]
For organisations, it is therefore important not only to look at the final model outcome. The origin, necessity and further use of personal data must also be assessed. A piece of data found is not automatically correct or suitable for every purpose. The AP advises the sector to map data use and test it against privacy legislation.
Three recommendations from the supervisory authority
The first recommendation is human intervention in decisions with major consequences. The second is awareness of data use: take stock of which personal data are processed and test the processing against the rules. The third is transparency towards debtors. Agencies must explain clearly how algorithmic systems are deployed and communicate proactively about the rights of data subjects. [1]
The AP considers it realistic that fully automated, prohibited decision-making occurs without agencies being aware of it. This is a risk assessment by the supervisory authority, not a finding that all organisations examined are breaking the law.
Consequences for practice
Careful implementation requires a defined role for AI, a responsible member of staff and a route for errors or objections. This also applies when a supplier offers a complete automatic workflow. The publication makes no statement about individual products, including Credimigo. Our Credimigo product profile should therefore be read separately from this supervisory news. For broader background: AI and supervision in the Netherlands.
The AP also announces a round-table discussion with the debt collection sector for 2 December 2026. This article provides general information and not legal advice. [1]
Sources
This article was produced with AI assistance from the original sources cited and checked against our editorial policy. Editorial policy