EU database registration

Registration is the only part of the high-risk regime that is public by design, which makes it a reputational and competitive surface as well as a legal one: your intended purpose statement and your Article 6(3) reasoning will be readable by customers, journalists and competitors. It is also a hard gate — registration must happen before market placement, so an unregistered system is unlawfully on the market no matter how complete its documentation is.

A provider decides under Article 6(3) that its Annex III system performs only a narrow procedural task and is therefore not high-risk. That decision does not remove it from the register: it must still register the system and the grounds for the conclusion, which means the reasoning has to be written to be read by an authority and by anyone else who looks — a far higher bar than an internal memo.

Assuming registration comes after launch, like a filing. It is a precondition of placing the system on the market or putting it into service, and it applies to Article 6(3) non-high-risk determinations too — a category many providers assume is exempt from every high-risk formality.

Is the EU AI database public?

Largely yes. The information registered is publicly accessible and user-friendly by design, with a restricted section for high-risk systems in law enforcement, migration, asylum and border control accessible only to the Commission and national authorities.

Do deployers have to register anything?

Deployers that are public authorities, Union institutions, or bodies acting on their behalf must register their use of high-risk AI systems. Private-sector deployers generally do not.

Do we register a system we decided is not high-risk?

Yes, where the system falls under Annex III and you concluded under Article 6(3) that it is not high-risk. Both the system and the grounds for that conclusion must be registered.