Article 7: Amendments to Annex III | EU AI Act

Article 7 of the AI Act elaborates on when and under what conditions the Commission can amend Annex III. These amendments can add and modify the list in Annex III by adding high-risk AI systems if both the following conditions are met:

• The AI system will be used in any of the areas mentioned in Annex III;
• The AI system poses a risk to the health, safety, or fundamental rights, that is, in respect of its severity and probability of occurrence, equal to or greater than the risk of harm posed by the high-risk AI system referred to in Annex III.

In cases where the Commission has to assess the aforementioned second condition, it must take into account the following criteria:

• The intended purpose of the AI system;
• The extent to which the AI system has been used or is likely to be used;
• The nature and amount of data processed and used by the AI system, especially special categories of personal data;
• The extent to which AI systems act autonomously and the possibility of humans overriding decisions and recommendations that may lead to adverse consequences;
• The extent to which the AI system has already caused harm to the health, safety, or fundamental rights and has given rise to significant concerns related to the materialization of such harm or adverse impact proven by reports or documented allegations that have been submitted to a national competent authority;
• The potential impact of such harm, particularly in terms of its intensity and ability to affect a plurality of persons or to disproportionately affect a particular group of persons;
• The extent to which the persons harmed are dependent on the outputs generated by the AI system, particularly for legal or practical purposes where opting out is not an option;
• The extent to which potentially harmed persons are in a vulnerable position in relation to the user of the AI system owing to an imbalance of power, knowledge, economic or social circumstances, or age;
• The extent to which the outputs generated by the AI system are reversible, whereby any outcomes that impact the health or safety of a person will not be considered as easily reversible;
• The extent the existing Union law provides for:
  • Effective measures to mitigate any risks posed by the AI system except for claims of damages;
  • Effective measures to prevent or minimize such risks.

Furthermore, the Commission is empowered to adopt the delegated acts per Article 97 to amend the list in Annex III by removing any high-risk AI systems when the following conditions are met:

• The high-risk AI system in question no longer poses a significant risk to the fundamental rights, health, or safety of users;
• The deletion will not result in a decrease in the overall level of protection of health, safety, and fundamental rights under Union law.