Identifying high-risk AI: Annex III and Article 6

When is an AI system high-risk?

Art. 6 of the AI Act sets out two routes:

1. Products (Art. 6(1), Annex I): The AI is a product or a safety component of a product that must be assessed by an independent body under EU law, for example under the Medical Device Regulation (MDR). These obligations apply from 2 August 2028.
2. Purposes of use (Art. 6(2), Annex III): The AI is used for one of the purposes listed there. Following the amendment by Regulation (EU) 2026/1744, these obligations apply from 2 December 2027.

What matters is the intended purpose: what the system is designed and used for. The same technology can be uncritical in one use and high-risk in another.

Which areas does Annex III list?

No.	Area	Examples from the Annex
1	Biometrics	remote identification, emotion recognition, categorisation by sensitive attributes
2	Critical infrastructure	safety components in road traffic, the supply of electricity, gas, water and heating, digital infrastructure
3	Education	admission, evaluation of learning outcomes, monitoring of exams
4	Employment	targeted job advertisements, screening and evaluating applications, promotion, termination, allocation of tasks, performance monitoring
5	Essential services	assessing, granting, reducing or revoking entitlement to public assistance benefits including healthcare services; creditworthiness; pricing of life and health insurance; classifying emergency calls, dispatching emergency services, triage in the emergency department
6	Law enforcement	risk assessment, evaluation of evidence, profiling
7	Migration, asylum, border control	examination of applications, risk assessment
8	Justice and democratic processes	assistance in interpreting the law, influencing elections

When does the exception under Art. 6(3) apply?

A system from Annex III is not considered high-risk if it does not pose a significant risk to health, safety or fundamental rights and does not materially influence the outcome of a decision. This is the case if it

• performs a narrow procedural task, such as sorting incoming mail by topic,
• improves the result of a previously completed human activity, such as polishing the language of a finished text,
• detects decision-making patterns or deviations without replacing or influencing the human assessment without review, or
• performs a preparatory task for an assessment, such as summarising documents.

Profiling of natural persons is always high-risk, even if one of these conditions applies. Anyone relying on the exception documents the assessment before placing the system on the market and registers the system (Art. 6(4), Art. 49(2)). The Digital Omnibus has shortened the information required for this registration.

How do you check your use?

1. Describe the purpose: What exactly does the AI do, and who uses the result for what?
2. Check Annex I: Is the AI part of a medical device or another regulated product? For hospitals: Is AI a medical device?
3. Check Annex III: Does the purpose fall within one of the eight areas?
4. Check the exception: Does a condition from Art. 6(3) apply, and is there no profiling?
5. Document the result: Even a “not high-risk” should be recorded with reasons in your AI inventory.

In May 2026 the Commission published draft guidelines with examples. They are a good aid, but not yet final.

What follows from the classification?

Providers of high-risk AI need, among other things, risk management, technical documentation, logging and a conformity assessment. Deployers must use the system in accordance with the instructions for use, assign human oversight to competent people, keep logs for at least six months and inform staff and affected persons (Art. 26). Public bodies and private entities providing public services also carry out a fundamental rights impact assessment.

How does this work with basebox?

basebox is a platform, not a high-risk system. Which obligations you have depends on the use. basebox is not built to assess people, carry out profiling or evaluate evidence; under the EU AI Act, that counts as high-risk. basebox makes no automated decisions: it makes suggestions, people make decisions. Every answer shows its sources with the quoted passage, and the specialist checks against the source. Your administrator decides which apps exist in your organisation and who may use them. More under Security.

Not legal advice: The classification depends on the specific purpose and has to be checked case by case. As of 11 October 2026; the Commission’s guidelines may still change the interpretation.