The proposed simplification of the AI Act is a balancing act between simplification, maintaining safeguards for high-risk AI, and responsible use.
On 7 May 2026, the Council and the European Parliament (EP) settled on a provisional agreement on the proposal to amend the EU AI Act. The proposal falls in the EU’s simplification agenda, which is comprised of ten omnibus packages proposed by the European Commission (EC). The proposed amendments are part of the “Omnibus VII: digital” legislative package, which seeks to simplify the digital legislative framework. The Digital Omnibus on AI was introduced in November 2025 with the purpose of simplifying the Artificial Intelligence (AI) Act, along with the Data Union Strategy. The amendments to the regulation were introduced as an attempt to response to implementation and compliance challenges of the AI Act by clarifying and harmonising the rules, seeking to ease out the administrative burden. Therefore, to provide sufficient time to ensure standards and support measures are in place, the agreement postpones the implementation agenda regarding the use of AI systems in high-risk areas, including education, to December 2027, as well as to August 2028 for the integration of AI into products.
Among the modifications proposed, few amendments aim to enhance innovation by simplifying the rules for businesses. The proposal highlights the importance of small mid-cap enterprises (SMC), which are enterprises that exceed the size of small and medium-sized enterprises (SME), in the EU’s economy. SMCs have a higher potential than SMEs to accelerate innovation and digitalisation, often proved through rapid growth. However, they face similar difficulties to SMEs regarding administrative burden; thus, necessitating greater support in the transition phase from SMEs into SMCs. The provisional agreement extends certain privileges granted to SMEs to SMCs, when relevant and while maintaining the objectives and protection criteria of the AI Act. For example, these exceptions allow simpler ways to provide certain compliance documents. Additionally, SMEs, start-ups, and SMCs are given priority access to regulatory sandboxes created at EU level by the AI Office. Opportunities to conduct real-world testing of AI systems are also further extended. In this context, the amendments clarify the AI Office’s role, and it is entrusted to oversight and facilitate compliance. Overall, these modifications reflect the need for simplification and reduction of the administrative burden to ensure that the EU can effectively foster innovation and competitiveness.
While these recent changes are primarily designed to support businesses and innovation, they also have important implications for specific sectors, most notably education, where the use of AI is subject to particularly stringent requirements. Under the EU AI Act, AI systems used in education are classified as high-risk if they determine access, admission, or evaluate learning outcomes. These systems face strict pre-market compliance obligations, including rigorous data governance, mandatory human oversight, and registration in an EU database before they can be deployed. The rules for high-risk systems will apply as of late 2027. Nevertheless, decision-makers continue a dialogue on how to balance the risks and opportunities, that come with AI in this domain: On 11 May, the Council approved conclusions that call for an ethical, safe, and human centred-approach of deploying AI in education systems, affirming that AI offers opportunities for learning and teaching which should be embraced within the EU. Above all, AI and digital skills for teachers and learners are recognised as important cornerstones of equitable access and autonomous technology use. The conclusions acknowledge that teachers need support in mastering the rapid changes that AI is introducing in education systems. To that end, the Council proposes several concrete measures, including the development of education-specific AI tools which are co-created by teachers. When such systems are developed on home turf, they can further support the EU’s strategic autonomy. The Commission is now expected to develop further recommendations in their forthcoming 2030 Roadmap on the future of digital education and skills.
Against this backdrop, the provisional agreement between the Council and the EP represents not only a technical adjustment of the AI Act, but also a broader effort to align simplification, innovation, and responsible use across sectors. As such, it marks another step in the EU’s ambition to enhance its competitiveness while positioning Europe as an attractive environment for both innovators and users of trustworthy AI.