What will the process be like to get a high-risk application approved by the EU commission?
The process for getting a high-risk AI application approved by the EU Commission is still under development, but it will likely involve the following steps:
Pre-assessment phase: Developers will need to conduct a self-assessment of their AI application to determine whether it falls within the scope of the AIA. If the assessment determines that the application is high-risk, the developer will need to prepare a conformity assessment report.
Notification phase: Developers will need to notify the relevant Notified Body of their intention to market or put into use their high-risk AI application. The Notified Body is an independent organization that is authorized by the EU Commission to assess conformity assessment reports.
Conformity assessment phase: The Notified Body will conduct a conformity assessment of the AI application. This assessment will typically involve reviewing the developer's conformity assessment report, conducting on-site inspections, and testing the AI application.
Authorization phase: If the Notified Body determines that the AI application complies with the AIA, it will issue an EU Declaration of Conformity. This declaration will state that the AI application has been assessed and found to meet the requirements of the regulation.
Marketing and use phase: Once the AI application has received an EU Declaration of Conformity, it can be marketed and put into use in the EU. Developers will need to maintain documentation of their conformity assessments and keep records of any changes made to the AI application. They will also need to comply with post-market surveillance requirements, which may involve submitting periodic reports to the EU Commission.
The AIA's conformity assessment requirements are designed to ensure that high-risk AI applications meet the necessary safety and performance standards before they are marketed or put into use. By requiring independent assessment of AI applications, the AIA aims to promote the development and deployment of safe, reliable, and responsible AI systems in the EU.