The EU AI Act has initiated its first amendment process just one year and eight months after taking effect, marking a rapid transition in EU AI governance from legislative completion to an implementation calibration phase.

This round of amendments is not a simple deregulation; rather, it is a parallel process of procedural relief combined with the targeted expansion of scope for new risks. The revision stems from implementation challenges, such as the absence of compliance support tools during the Act's rollout, and the practical necessity of the EU's Simplification Revolution to address slowing productivity, intensifying global competition, and the need to boost industrial competitiveness.

A consensus has been reached among EU Trilogue institutions to extend the compliance deadline for high-risk AI systems and expand compliance facilitation measures to small and mid-cap companies. At the same time, they have refused to lower the threshold for fundamental rights protection in the name of regulatory relief. This includes restoring registration obligations for high-risk systems, maintaining the strictly necessary standard for sensitive data processing, and adding new prohibitions against non-consensual generative AI content (deepfakes) to address emerging technical risks.

This amendment is a targeted adjustment addressing implementation issues rather than a reconstruction of regulatory philosophy. While balancing the reduction of corporate burdens, the EU remains steadfast in its core regulatory principles, refusing to compromise the bottom line of rights protection due to external competition.

Translated by Yongjia LIN

Full text in Chinese available at:

https://fddi.fudan.edu.cn/d4/48/c18965a775240/page.htm