UPDATE – JUNE 2025: The EU AI Act officially entered into force on August 1, 2024, with key provisions in the timeline taking effect in stages through 2027. Bans on unacceptable-risk AI practices and AI literacy requirements are already enforceable as of February 2025. However, most high-risk AI system obligations will not be fully in force until August 2026–2027, pending finalization of technical standards and supporting guidance. The Act remains the world’s most comprehensive AI law, influencing legislation globally.
ORIGINAL NEWS STORY:
The European Union (EU) achieved a historic milestone on December 8, 2023, with the European Parliament and Council reaching a provisional agreement on the AI Act. According to an official press release, the Act is designed to ensure the safety of AI systems used in the EU, uphold fundamental human rights, and foster innovation. Key rules have been established to address potential risks associated with AI applications.
Banned Applications
The AI Act will explicitly prohibit several concerning AI applications, including:
Biometric categorization using sensitive characteristics.
Scraping facial images to create databases.
Emotion recognition in work and education settings.
Social scoring based on behavior.
AI that manipulates human behaviors.
AI that exploits vulnerabilities.
Law Enforcement Exemptions
Narrow exceptions permit biometric identification in public spaces for law enforcement purposes, provided there is judicial authorization. “Real-time” use is strictly limited to searching for victims, preventing terrorism, or identifying suspects in specific crimes. “Post-remote” use is restricted to searches related to serious crime victims.
High-Risk System Obligations
For AI systems deemed high-risk, specific obligations will be imposed regarding fundamental rights assessments, transparency, and oversight. Citizens will have the right to file complaints and receive explanations concerning decisions made by high-risk AI systems that affect their rights.
General AI System Requirements
General AI systems and models must adhere to transparency requirements, including technical documentation, copyright compliance, and detailed summaries of training content. High-impact models with systemic risk face additional obligations, covering risk assessments, testing, incident reporting, cybersecurity, and energy efficiency.
Innovation Support
Provisions within the AI Act aim to promote regulatory sandboxes for AI systems and models, fostering real-world testing to assist businesses, particularly small and mid-size enterprises, in developing AI solutions.
Fines
Systems and models found in violation of the AI Act will face fines ranging from 35 million Euros or 7% of turnover down to 7.5 million Euros or 1.5%, based on the severity of the infringement and the size of the company.
As for the next steps for the AI Act, the agreed text will now undergo formal adoption by both Parliament and Council to become EU law. The earliest the AI Act could take effect is 2025.
For assistance in how the EU AI Act timeline, and other bills around the globe, could impact your company when it comes to AI compliance, reach out to BABL AI. Their team of audit experts will offer valuable guidance.
