EU Agrees to Delay High-Risk AI Compliance Deadlines, Offering Businesses More Preparation Time

Overview of the Provisional Agreement

In a significant development for artificial intelligence regulation, European Union lawmakers reached a provisional agreement early Thursday to extend the compliance deadlines for high-risk AI systems under the AI Act. This move provides enterprises with additional time to align with the stringent requirements, easing the initial pressure of the original timetable.

The deal, struck between negotiators from the European Parliament and the European Council, introduces revised deadlines: December 2, 2027, for standalone high-risk AI systems, and August 2, 2028, for AI integrated into products governed by EU sectoral safety rules. The original enforcement date was set for August 2, 2026. These changes are outlined in an official statement from the European Parliament.

However, the agreement remains provisional until formally adopted by both the Parliament and the Council, a process expected to be completed before August 2. Until then, the original 2026 deadline technically applies as initially drafted.

Political and Economic Context

Marilena Raouna, Cyprus’s deputy minister for European affairs and representative of the EU Council presidency, highlighted the benefits: “Today’s agreement on the AI Act significantly supports our companies by reducing recurring administrative costs.” Cyprus currently holds the rotating presidency of the Council, which negotiates on behalf of all 27 member states. The breakthrough came just nine days after previous negotiations collapsed, underscoring the urgency and complexity of finalizing the regulatory framework.

Key Changes in Regulation and Compliance

The provisional agreement introduces several refinements to reduce regulatory overlap and provide clearer guidance for businesses. One major change removes duplicative requirements for AI systems in machinery products, which will now follow only sectoral safety rules with equivalent health and safety safeguards.

The definition of a “safety component” under the AI Act has been narrowed. AI features that merely assist users or enhance performance will no longer automatically be classified as high-risk, provided their failure does not create health or safety risks. This clarification, as noted in the Parliament’s statement, helps businesses avoid unnecessary compliance burdens.

Sectoral Overlap Resolution

For broader industries—including medical devices, toys, lifts, machinery, and watercraft—the co-legislators agreed on a mechanism to resolve overlaps between the AI Act and existing sectoral laws. This approach ensures coherence without compromising protection standards. The Council confirmed this in its own statement, emphasizing the goal of streamlined regulation.

New Timelines for Sandboxes and Watermarking

The deadline for member states to establish AI regulatory sandboxes has been extended by one year, now set for August 2, 2027. These sandboxes allow companies to test AI systems under regulatory supervision before full deployment.

Conversely, watermarking obligations for AI-generated content will apply sooner than originally proposed by the European Commission. Starting December 2, 2026 (instead of February 2, 2027), providers must label synthetic content to improve transparency and traceability.

Regulatory Sandboxes and Mid-Size Firms

Small mid-cap companies—those not qualifying as small or medium-sized enterprises—now gain access to exemptions previously reserved for SMEs. This expansion offers more breathing room for a broader range of businesses. Additionally, the deal clarifies that the EU’s AI Office will centrally supervise general-purpose AI systems, while national authorities retain responsibility for law enforcement, border management, judicial authorities, and financial institutions.

Political Reaction and Next Steps

Arba Kokalari, the Parliament’s co-rapporteur for the Internal Market and Consumer Protection committee, remarked: “With this agreement, we show that politics can move just as quickly as technology.” The statement reflects the ambition to keep pace with rapid AI advancements while ensuring robust governance.

The provisional deal still requires formal adoption by both the European Parliament and the Council. Once approved, it will replace the original deadlines and become binding law. Stakeholders are advised to monitor the legislative process closely, as the current Aug. 2, 2026 deadline remains in effect until adoption.