The European Union (EU) has reached a provisional agreement on the AI Act, a comprehensive legal framework that restricts the use of artificial intelligence. Ursula von der Leyen, the President of the EU Commission, hailed it as the world’s first complete legal framework on Synthetic Intelligence. The agreement emphasizes quantifiable risks, providing constitutional certainty for the sector and technological advancement.
The AI Act introduces a tiered risk control system, categorizing AI techniques based on their impact on rights. Most AI methods are classified as “minimal chance,” covering applications like spam filters and auto recommendation systems. Service providers will need to adhere to Artificial codes of conduct.
Stricter regulations will apply to AI systems labeled as “high-risk,” such as those used in vital infrastructures, education evaluation, law enforcement, or biometric identification. Compliance may involve risk-mitigation measures, human oversight, high-quality datasets, and thorough documentation.
Instances that pose a threat to fundamental rights fall under the “unacceptable risk” category and are prohibited. Examples include forecast policing, workplace mental recognition systems, and discriminatory practices based on traits like race or sexual orientation.
Deepfakes and AI-generated content must be appropriately labeled, and users interacting with AI systems should be informed of the system’s nature. Additionally, foundational AI models requiring extensive training may face additional regulations after a year.
Violating the AI Act can lead to significant fines, ranging from 1.5% to 7% of a company’s global turnover or specified amounts. The Act allows for different fine structures for SMEs and startups.
Barry Scannell, an AI law expert, highlighted the economic risks of non-compliance with the Act. He emphasized the need for transparency, data quality improvement, and bias management tools, which might increase operational costs for businesses.
After 38 days of negotiations, the agreement was signed, with concerns raised by European Parliament member Svenja Hahn regarding creativity, legal rights, and potential overregulation. The Act received mixed reactions from Big Tech representatives, with the Computer and Communications Industry Association (CCIA) expressing concerns about its impact on future AI policy and technology advancement in Europe.
The agreement awaits formal approval from the European Parliament and the Council before coming into effect. Ursula von der Leyen has committed to advancing AI regulation globally through participation in various international organizations.
