<center>Council of the European Union and European Parliament, 7 May 2026</center>
The provisional agreement forms part of the EU's wider simplification agenda, which has seen the Commission put forward ten "Omnibus" packages since February 2025 aimed at reducing administrative, regulatory and reporting burdens across a range of sectors. The AI-specific package, the <span class="news-text_italic-underline">Digital Omnibus on AI</span>, was first proposed by the Commission on 17 November 2025 and has been treated as a legislative priority by both institutions, given that provisions on high-risk AI systems are due to enter into force on 2 August 2026.
The simplification agenda itself was set in motion by the European Council in October 2024, informed by the Letta and Draghi reports on European competitiveness and further propelled by the Budapest Declaration of November 2024, which called for a "simplification revolution" in EU regulation.
<span class="news-text_medium">Revised timelines for high-risk AI systems:</span> Rather than adopting the Commission's proposed flexibility of up to 16 months, the co-legislators have introduced fixed revised application dates: 2 December 2027 for stand-alone high-risk AI systems and 2 August 2028 for high-risk AI systems embedded in products.
<span class="news-text_medium">Child protection and prohibited practices:</span> A new provision has been added to the AI Act expressly prohibiting AI practices that generate non-consensual sexual and intimate content or child sexual abuse material, marking a notable strengthening of the Act's prohibited practices chapter.
<span class="news-text_medium">SME and SMC exemptions:</span> Certain regulatory exemptions previously available only to small and medium-sized enterprises are extended to small mid-caps, reducing the compliance burden on a broader category of businesses.
<span class="news-text_medium">EU database registration:</span> The provisional agreement reinstates the obligation for providers to register AI systems in the EU database for high-risk systems, even where those providers consider their systems to be exempt from high-risk classification.
<span class="news-text_medium">Sensitive personal data processing:</span> The standard of strict necessity for processing special categories of personal data for bias detection and correction purposes has been reinstated, reversing a proposed relaxation.
<span class="news-text_medium">AI regulatory sandboxes:</span> The deadline for competent national authorities to establish AI regulatory sandboxes has been extended to 2 August 2027.
<span class="news-text_medium">Transparency obligations:</span> The grace period for providers to implement transparency solutions for artificially generated content has been reduced from six months to three months, with the new deadline set at 2 December 2026.
<span class="news-text_medium">AI Office competences:</span> The agreement clarifies the supervisory competences of the AI Office in relation to AI systems based on general-purpose AI models, where the model and the system are developed by the same provider. Exceptions are listed for areas where national authorities retain competence, including law enforcement, border management, judicial authorities and financial institutions.
<span class="news-text_medium">Sectoral legislation and overlap:</span> A compromise mechanism has been agreed to resolve situations where sectoral legislation — covering sectors such as medical devices, toys, lifts, machinery and watercraft, contains AI-specific requirements similar to those in the <span class="news-text_italic-underline">AI Act</span>. Implementing acts will limit the <span class="news-text_italic-underline">AI Act's</span> application in such cases to avoid duplication. Separately, the machinery regulation has been exempted from direct applicability of the AI Act, with the Commission empowered to adopt delegated acts adding health and safety requirements for high-risk AI systems under that regulation. The Commission is also required to provide guidance to assist economic operators in minimising their compliance burden under both the AI Act and applicable sectoral legislation.
The provisional agreement must be formally endorsed by both the Council and the European Parliament before undergoing legal and linguistic revision, with formal adoption expected in the coming weeks.
This agreement represents the first deliverable under the "One Europe, One Market" roadmap agreed by the EU institutions and signals a pragmatic recalibration of the <span class="news-text_italic-underline">AI Act's</span> implementation timeline in response to industry concerns about readiness. For businesses operating high-risk AI systems, particularly those in regulated sectors, the fixed revised application dates provide much-needed certainty. The reinforced child protection provisions and the clarified AI Office competences are also significant developments that practitioners advising on AI governance and compliance should note carefully.
