AI Act: the Commission clarifies the classification of high-risk AI systems

Section 6 of the IA Act

As a reminder, Article 6 sets out the rules for classifying AI systems as high-risk systems. In particular, the AI Act identifies two types of “high-risk” AI systems:

• AI systems incorporated into products regulated by the EU’s harmonized legislation on product safety
• AI systems that are likely to have a significant impact on the health, safety, or fundamental rights of individuals in specific use cases listed in the AI legislation

The Commission has published draft guidelines on the classification of so-called “high-risk” AI systems referred to in Article 6 of the Regulation.

Key points

1. The guidelines assist providers and deployers of AI systems in assessing whether their AI system poses a high risk

2. They provide a non-exhaustive list of examples covering a wide variety of fields and use cases, which may be updated by the Commission; here are a few examples related to the administration of justice:

• video surveillance cameras installed in the walls and/or ceilings of a stadium: If an incident occurs during a match, the captured biometric data (facial images) is compared against a biometric database to identify the perpetrators, without the active participation of stadium visitors. This AI system is classified as high-risk
• A video surveillance system with facial recognition installed at a stadium entrance to detect individuals on a blacklist who are barred from entering the stadium through direct access control: Individuals must actively participate – that is, present themselves to the AI system – to gain access to the stadium; therefore, this AI system would not be classified as high-risk

Implementation and Enforcement Deadlines

The guidelines confirm the postponement of the effective dates for the relevant provisions of the AI Act. Specifically:

• The rules applicable to systems used in certain high-risk areas – including biometrics, critical infrastructure, education, employment, migration, asylum, and border control – take effect on December 2, 2027.
• For systems embedded in products such as robotics and industrial machinery, the rules apply as of August 2, 2028.

Expert opinion

Although non-binding, the Commission’s guidelines effectively dictate the enforcement and sanctioning policy that would be applied by regulatory authorities.

Firsh supports companies in conducting an audit of their AI solutions and assists them in classifying them under the AI Act.