EU Lawmakers Clash Over Foundation Models in Final AI Act Push

European institutions face intense pressure to finalize the EU AI Act before the end of 2023, but deep disagreements over foundation models threaten to delay or dilute the landmark legislation.

European lawmakers currently engage in intense Trilogue negotiations to finalize the EU AI Act, facing mounting pressure to pass the legislation before the end of 2023. With European elections looming in June 2024 and competing AI regulatory frameworks emerging in the US and China, the EU pushes hard to establish its first comprehensive rules for artificial intelligence. However, as the finish line approaches, significant controversies among the Parliament, Commission, and Council become increasingly apparent.

The most prominent friction point centers on the regulation of foundation models, highlighting a major divide between the Council's and the Parliament's approaches. This resurfacing disagreement forces the addition of extra negotiation sessions to the already packed schedule. Lawmakers struggle to find a middle ground on how strictly to govern these powerful, general-purpose AI systems amidst concerns about stifling innovation versus preventing societal harm.

Because of the pressing urgency to secure a political compromise before the year concludes, these growing divergences likely result in the removal or watering down of the most contested elements. Instead of embedding these highly debated rules directly into the primary text, lawmakers plan to defer the complex details to the proposed AI Office or future standardization initiatives. This strategic shift allows the EU to claim a timely legislative victory while leaving the hardest regulatory questions for later.

Read More at the original source →