Algorithm registers: what does a public body publish, and when?
Short answer
A public algorithm register shows which high-impact algorithms a public body uses: what it does, for what, with which data and who is responsible. It sits alongside two other lists: the EU database, where public bodies register their high-risk AI under the AI Act, and your own internal AI register, which holds everything, including what is never published.
Three lists, three purposes
| List | What belongs in it | For whom |
|---|---|---|
| Public algorithm register | High-impact algorithms, with purpose, workings, data and contact | Residents, journalists, elected representatives |
| EU database (AI Act) | High-risk AI deployed by a public body | Supervisors and the public |
| Your own AI register | Everything the organisation uses, including the writing assistant | CISO, DPO, management |
Without that third list you cannot fill the first two: you simply do not know what is running. See Setting up an AI register.
What counts as high-impact?
No button gives the answer. In practice, four questions help:
- Does it affect residents or businesses? An algorithm that helps decide an application, an inspection or a tariff affects people directly.
- Is it decisive? A system whose outcome is nearly always followed is decisive, even if there is formally a person in between.
- Is it about people? Profiling, ranking or selecting people calls for openness.
- Would you want to explain it in public? If the answer is hesitant, it probably belongs in the register.
An internal writing assistant usually falls outside this. A model that prioritises reports or ranks applications falls within.
What goes into an entry?
- the purpose and expected workings, in plain language;
- which data is used and where it comes from;
- how human oversight is arranged and who intervenes;
- which trade-offs and risks were identified, and which measures were taken;
- who is responsible and where people can go with questions.
For high-risk AI, the AI Act adds a fundamental rights impact assessment and registration in the EU database. See The EU AI Act in 2026.
Keeping it current is the real work
A register ages the moment a model is replaced, a source changes or an automation is extended. So tie the publication to your own operations: whoever changes a model or source also updates the entry.
With werqly that information comes from the system itself. In werqly AI Control it is recorded per role which models, sources and automations are released, and the audit log shows what was actually used. That saves asking around departments. See AI for local government and Generative AI in local government.
Frequently asked questions
Questions on this topic
Does every AI application belong in an algorithm register?
No. Such registers cover high-impact algorithms: applications with consequences for residents or businesses. An internal writing assistant usually does not belong there, but it does belong in your own AI register, so you know what is running.
How does this differ from the EU database in the AI Act?
A national algorithm register, such as the Dutch one, is a transparency facility for high-impact algorithms. The EU database is an AI Act obligation: public bodies register their use of high-risk AI there. An application can belong in both.
Who fills in the register?
Usually the owner of the application, supported by the CISO, the data protection officer and information management. Agree who updates the entry when something changes, or the register quietly goes out of date.
Does an AI environment of your own help with the register?
Yes, because the data comes from the system: which model, which sources, which roles and what use. That is faster and more reliable than asking around departments, and it stays current when something changes.
Sources
This article is general information, not legal advice. Rules and guidance change; check the sources listed or ask your lawyer or DPO if in doubt.