Risk classification

There are four outcomes, and they are not a spectrum of severity so much as four different legal regimes.

Prohibited practices under Article 5 cannot be deployed at all, at any level of care. These include social scoring by public authorities, untargeted scraping of facial images, emotion recognition in workplaces and educational institutions, exploitation of vulnerabilities of specific groups, and certain predictive policing based solely on profiling. These provisions have applied since 2 February 2025.

High-risk systems are defined by Article 6. A system is high-risk if it is a safety component of a product covered by the Annex I legislation, or if it falls within one of the eight areas in Annex III: biometrics, critical infrastructure, education and vocational training, employment and worker management, access to essential private and public services including credit scoring, law enforcement, migration and border control, and the administration of justice.

Limited-risk systems carry transparency duties under Article 50: people must be told they are interacting with an AI, and synthetic audio, image, video or text content must be marked as artificially generated. Minimal-risk systems carry no specific obligations beyond voluntary codes.

Article 6(3) provides a narrow derogation: a system in an Annex III area is not high-risk if it does not pose a significant risk of harm, for instance because it performs a narrow procedural task. Relying on that derogation requires you to document your assessment, which is precisely why the platform records reasoning rather than just outcomes.

The assessment presents the Annex III criteria as structured questions covering intended purpose, the people affected, the system's autonomy, and the potential for harm to health, safety or fundamental rights.

It produces a classification with the answers retained alongside it, so the result is traceable to the reasoning that produced it.

You can override the outcome where you have grounds. The override, its author and its justification are all recorded. An override is a defensible judgement only if it is documented as one.

Classifying a system as high-risk automatically generates its full obligation set: the AI Act's Articles 9 through 15, plus registration, conformity assessment and post-market monitoring.

A Prohibited outcome is enforced, not just recorded: the platform automatically raises a critical compliance action naming the system, and blocks that system's deployment status from being set to production anywhere in the platform until the classification changes.