First Requirements of the AI Act Come Into Effect
On February 2, 2025, the first requirements under the European Union AI Act officially came into effect. This landmark legislation includes a ban on the use of AI systems that involve prohibited AI practices and mandates that providers and users of AI systems ensure they possess sufficient AI literacy to effectively operate these technologies.
Pending Guidance from the European Commission
Despite the urgency surrounding the implementation of the Act, guidance from the European Commission is still pending. This has led many companies to independently design and implement their own strategies and compliance plans in the absence of formal directives.
Engagement at Knowledge Sharing Forums
In light of the new regulations, law firms are increasingly engaging in knowledge sharing forums. For instance, Clifford Chance is preparing for the global AI Action Summit in Paris. During the satellite event known as the AI Fringe on February 11, Dessislava Savova, a partner at Clifford Chance, will moderate a discussion focused on delivering trustworthy AI during challenging times. The panel will feature notable experts, including Brendan Kelleher from SoftBank and Laurent Daudet from LightOn.
Timeline of the EU AI Act
The EU AI Act was officially enforced on August 1, 2024, with its requirements being implemented over a staggered timeline. Most provisions are scheduled for implementation by August 2, 2026, affecting any systems utilized within the EU.
Literacy Requirements
The Act stipulates that providers and deployers of AI systems must take appropriate measures to ensure that their staff possess adequate skills, knowledge, and understanding of AI systems. It is crucial that these individuals are aware of the potential risks and harm that AI can cause.
Prohibited AI Practices
Prohibited AI practices are defined as those that are considered harmful and abusive, contradicting the values of the Union, the rule of law, and fundamental rights. The Act bans systems that:
- Are manipulative or deceptive
- Exploit vulnerabilities, such as age or socio-economic status
- Score individuals based on behavior or personality traits
- Profile individuals to predict criminal behavior
- Create or expand facial recognition databases through untargeted scraping of the internet and CCTV footage
- Infer emotions or categorize individuals based on race or political beliefs
Further Reading
For more insights on the bans that came into effect, a paper from Mayer Brown provides additional context and analysis of the EU AI Act and its implications.
