Supervision and Enforcement of the European Union’s AI Act
The European Union’s Artificial Intelligence Act (AI Act) aims to establish a regulatory framework for the oversight and enforcement of AI technologies across member states. This article provides an in-depth overview of the authorities involved in the supervision and enforcement of the AI Act, alongside potential penalties for noncompliance.
Authorities and Other Relevant Actors at EU Level
At the EU level, the AI Act introduces several key authorities responsible for ensuring compliance and effective implementation of the regulations.
AI Office
The AI Office, established within the European Commission, plays a central role in the development of compliance tools and reporting obligations for AI systems. Key responsibilities include:
- Development of Compliance Tools: The AI Office is tasked with creating model terms for contracts between high-risk AI system providers and third parties, as well as templates for fundamental rights impact assessments.
- Information About General-Purpose AI Models: Providers of general-purpose AI models must submit necessary documentation to demonstrate compliance with the AI Act.
- Reporting Obligations: Providers of AI models that pose systemic risks must report serious incidents and corrective measures to the AI Office.
- Regulatory Sandboxes: National authorities must inform the AI Office about regulatory sandboxes and may seek its guidance.
- Support for SMEs and Startups: The AI Office will provide standardized templates to aid small and medium-sized enterprises in complying with the AI Act.
European AI Board
The European AI Board is another key entity, advising the Commission and EU Member States on consistent application of the AI Act. Its composition includes one representative from each EU Member State, alongside the AI Office and the European Data Protection Supervisor.
The Board’s primary mission encompasses:
- Facilitating coordination between national authorities and harmonizing administrative practices.
- Assisting in the development of organizational and technical expertise necessary for the AI Act’s implementation.
- Issuing recommendations on the effective application of the AI Act.
Advisory Forum
An Advisory Forum will provide technical expertise and advise both the Board and the Commission. It comprises a diverse range of stakeholders, ensuring balanced representation from industry and civil society.
Tasks of the forum include:
- Preparing opinions and recommendations for the Board or Commission.
- Publishing annual reports on its activities.
Scientific Panel of Independent Experts
A Scientific Panel of Independent Experts will support the AI Office, consisting of experts selected for their technical expertise in AI. The panel’s role includes:
- Advising the AI Office on potential systemic risks.
- Contributing to the development of evaluation tools for AI models.
Authorities and Other Relevant Actors at National Level
At the national level, each EU Member State is required to establish or designate specific authorities responsible for enforcing the AI Act:
Market Surveillance Authorities
Each member state must designate at least one Market Surveillance Authority to ensure compliance with the AI Act and enforce its provisions.
Notifying Authorities
Member States must also establish Notifying Authorities to oversee conformity assessment bodies responsible for third-party testing and certification of AI systems.
Guidance and Advice
Market surveillance authorities will provide guidance, particularly to SMEs and startups, on implementing the AI Act while considering advice from the Board and the Commission.
Sufficient Resources
Member States are obligated to ensure that market surveillance authorities are equipped with adequate resources to fulfill their roles effectively. Reports on the status of these resources must be submitted to the Commission regularly.
Penalties
Under the AI Act, EU Member States are required to define rules for penalties applicable to infringements. These penalties must be:
- Effective: Ensuring compliance with the regulations.
- Proportionate: Reflecting the severity of the infringement.
- Dissuasion: Serving as a deterrent against future violations.
Penalties for noncompliance include:
- Administrative fines of up to €35 million or 7% of global revenue for violations related to prohibited AI systems.
- Fines of up to €15 million or 3% of revenue for breaches concerning limited and high-risk AI.
- Fines of up to €7.5 million or 1% of revenue for providing incorrect or misleading information to authorities.
This regulatory framework aims to create a safe and compliant environment for the deployment of AI technologies across the EU, balancing innovation with the necessary oversight to protect public interests.
