Europe Enforces AI Rules: A Challenge for American Giants

The EU Implements Its AI Law

European legislation on artificial intelligence enters a crucial phase this week. Starting August 2, regulators from the European Union, in collaboration with national authorities, began enforcing rules that require companies to clearly label AI-generated content. Additionally, they must inform users when they are interacting with AI technology rather than a human.

Transparency and Enhanced Obligations

The new measures introduce transparency obligations for general-purpose AI models, such as chatbots. They also prohibit certain practices deemed unacceptable. Companies must now clearly indicate to users when they are interacting with AI. Furthermore, any content generated or modified by AI, including deepfakes, must be appropriately labeled. Some content even requires machine-readable markers to facilitate identification.

Oversight and Rule Enforcement

The AI office of the European Commission is responsible for overseeing providers of general-purpose AI models. Meanwhile, national regulators are tasked with enforcing the rules for the majority of other AI systems. This division aims to ensure effective and consistent enforcement of the new standards across Europe.

Implications for Businesses

Marcus Evans, a partner at the law firm Norton Rose Fulbright, emphasizes that the transparency requirements of the AI law are often underestimated by organizations. He clarifies that these rules apply not only to large AI developers but also to companies that use AI tools under their own brand. Evans recommends that businesses work closely with their technology providers to ensure compliance with the new obligations.

Financial Penalties and Pressure on American Companies

Companies that fail to comply with these new rules risk fines of up to 3% of their annual global revenue. These potential penalties increase the pressure on American AI companies serving clients in Europe. Peter Van Dyck, a partner at the law firm A&O Shearman, warns that American AI labs must be aware that their models, if they reach European users, subject them to these new rules.

A Gradual Enforcement of Rules

Although regulators have the power to impose severe penalties, Van Dyck anticipates a gradual enforcement of the rules. Companies will initially need to demonstrate that they are actively striving to comply with the new requirements. This phased approach aims to encourage broader and faster adoption of practices compliant with European legislation.