The European Artificial intelligence Act: what you need to know
The European Artificial Intelligence Act marks a significant milestone in AI regulation. After extensive negotiations, the Council presidency and the European Parliament’s negotiators have reached a provisional agreement on harmonized rules for artificial intelligence (AI). The primary objective of this act is to ensure the safety of AI systems in the European market while upholding fundamental rights and EU values.
The Act introduces a risk-based approach, meaning that stricter regulations will apply to AI systems with higher potential risks.
It's the world’s first legislative proposal of its kind, positioning itself as a potential global standard for AI regulation, similar to the GDPR's role in data protection.
Key components of the provisional agreement include:
Classification of AI systems based on risk levels, distinguishing between high-risk AI and prohibited AI practices. Varying degrees of obligations are imposed, whereby high-risk AI systems undergo stringent requirements to access the EU market.
Enhanced governance at the EU level, prohibiting certain applications like cognitive behavorial manipulation, untargeted scraping of facial images, emotion recognition in sensitive contexts, social scoring and biometric categorization to infer sensitive data. Exceptions for law enforcement are foreseen.
Comprehensive measures for high-risk AI systems, mandating a fundamental rights impact assessment before deployment.
The Act establishes a new governance architecture, creating an AI Office within the European Commission to oversee advanced AI models, backed by a scientific panel of independent experts. Additionally, it defines penalties for violations, distinguishing between different types of infringements.
Transparency and fundamental rights protection remain central, with provisions mandating fundamental rights impact assessments and increased transparency for high-risk AI systems.
In support of innovation, the Act introduces measures such as AI regulatory sandboxes for controlled testing and validation, including provisions for smaller companies to alleviate administrative burdens.
While the provisional agreement sets the stage for adopting this transformative AI Act, further technical work is required before finalizing and implementing it across member states.
