The European Parliament and the Council of the EU reached a provisional political agreement on 7 May 2026 on a proposal to streamline rules under Regulation 2024/1689 laying down harmonised rules on artificial intelligence, the AI Act. The proposal forms part of the Omnibus VII legislative simplification package. Formal adoption by both institutions is required before the text becomes law. Co-legislators intend to complete adoption before 2 August 2026, when existing high-risk AI system obligations take effect under the current text.
The agreement amends Regulation 2024/1689. The new application date for stand-alone high-risk AI systems is 2 December 2027; for high-risk AI systems embedded in products, 2 August 2028. Regulatory exemptions available to small and medium-sized enterprises are extended to small mid-cap companies. Permitted processing of sensitive personal data for bias detection and mitigation is broadened. The powers of the European AI Office are reinforced and governance fragmentation across national competent authorities is reduced. A new prohibition under Article 5 covers AI systems that generate child sexual abuse material or non-consensual intimate images of identifiable persons. The deadline for establishing AI regulatory sandboxes at national level is extended to 2 August 2027.
Providers of stand-alone high-risk AI systems gain a compliance window extended to 2 December 2027 before mandatory conformity assessment and registration obligations apply. Manufacturers embedding high-risk AI in regulated products gain until 2 August 2028. Small mid-cap companies gain access to SME conformity assessment exemptions, reducing their pre-market obligations. Providers of AI systems capable of generating non-consensual intimate imagery must withdraw or reconfigure those systems regardless of deployment model. The European AI Office's reinforced authority directly affects providers of general-purpose AI models under Chapter V of the AI Act.
The provisional agreement is not yet law and may change before formal adoption. Obligations on general-purpose AI model providers under Chapter V of the AI Act are not addressed by this proposal and remain on their existing schedule. Member states that have already begun establishing national AI regulatory sandboxes should assess how the new August 2027 deadline applies to their implementation steps.
Licentium advises operators, providers, importers, and deployers of AI systems on AI Act obligations and compliance readiness. We may be able to assist directly or identify appropriate support through our partner network. Work we undertake includes AI Act conformity assessment support, high-risk system classification, general-purpose AI model compliance under Chapter V, provider obligations under Article 16, and AI governance documentation for regulated sectors.