EU Parliament Votes to Delay AI Act's High-Risk Compliance Deadlines to 2027

On June 16, 2026, the European Parliament voted overwhelmingly to delay the most stringent enforcement deadlines of the landmark EU AI Act. By a margin of 423 to 57, lawmakers approved the "Digital Omnibus on AI," a legislative package that fundamentally rewrites the compliance timeline for the world's first comprehensive artificial intelligence law.[1]

The vote averts what the European tech sector had warned would be a catastrophic regulatory cliff edge. Under the original text passed in 2024, strict requirements for "high-risk" AI systems were scheduled to take legal effect on August 2, 2026.[1][6]

The newly approved Omnibus pushes that deadline back by 16 months. Standalone high-risk systems—such as AI used in employment screening, credit scoring, and biometric identification—now have until December 2, 2027, to comply with the Act's rigorous auditing and governance mandates.[6][7]

For high-risk AI systems embedded as safety components in regulated physical products, such as medical devices, toys, or elevators, the delay is even longer. Those systems are now granted a reprieve until August 2, 2028, to align with broader sectoral safety regulations.[1][8]

The evidence supporting the delay centers on a severe logistical bottleneck within the EU's own regulatory apparatus. The original AI Act required companies to adhere to harmonized technical standards to prove their systems were safe. However, those standards had not yet been finalized by European standardization bodies, leaving the industry in a state of legal limbo.[4]

The political momentum for the delay was catalyzed by the September 2024 Draghi Report on EU competitiveness. The report delivered a stark diagnosis of the widening productivity gap between the European Union and the United States, attributing the shortfall largely to the tech sector and warning that over-regulation was stifling domestic innovation.[4][5]

Following the Draghi Report, the European Commission introduced the Digital Omnibus in November 2025 as a horizontal simplification effort. The goal was to eliminate overlapping regulations and reduce the compliance burden on European businesses attempting to catch up with American and Asian rivals.[2][8]

The most fiercely contested provision during the negotiations was the treatment of the industrial machinery sector. Germany, home to engineering giants like Siemens, lobbied aggressively to exempt industrial AI applications from the AI Act's scope, arguing that machinery was already heavily regulated by existing safety directives.[3]

The final text reflects a significant victory for the German industrial base. The machinery sector has been largely carved out of the AI Act's overlapping requirements, meaning industrial AI products will only need to comply with updated sectoral safety rules rather than facing double-jeopardy enforcement from the AI Office.[1][3]

The Omnibus also expands regulatory relief for smaller businesses. Previously, exemptions and simplified technical documentation requirements were reserved strictly for Small and Medium-Sized Enterprises (SMEs). The new legislation extends these privileges to "small mid-caps" (SMCs)—companies with up to 500 employees.[8]

While the legislation relaxes timelines for corporate compliance, it simultaneously accelerates enforcement against specific consumer harms. The Omnibus introduces a strict new prohibition on AI systems that generate non-consensual intimate imagery—often referred to as "nudifiers"—and child sexual abuse material.[1][6]

Providers and deployers of these malicious deepfake systems face an accelerated compliance deadline. They must implement adequate technical safeguards to prevent the generation of such material, or remove the systems from the EU market entirely, by December 2, 2026.[1]

Transparency mandates have also been placed on a faster track than the high-risk rules. Obligations requiring companies to add machine-readable watermarks to AI-generated content will take effect on December 2, 2026, granting companies a brief grace period to update their content provenance protocols.[1][8]

Despite the Parliament's decisive vote, a critical procedural hurdle remains. The Digital Omnibus is not yet binding law. It must be formally adopted by the European Council and published in the Official Journal of the European Union before August 2, 2026.[6]

If the Council fails to formalize the text before that date, the original AI Act provisions will automatically trigger, plunging the European tech sector into the exact regulatory chaos the Omnibus was designed to prevent. Legal experts advise organizations to maintain dual-track compliance preparations until the ink is officially dry.[6][7]

The compromise has drawn sharp criticism from privacy advocates and civil rights groups. Dutch Member of the European Parliament Kim van Sparrentak publicly condemned the delay during the trilogue negotiations, stating that "Big Tech is probably popping champagne" while companies that proactively prepared for the original safety rules are left at a disadvantage.[5]

Conversely, industry groups argue the delay was the only pragmatic path forward. Without the finalized technical standards, they contend, the August 2026 deadline would have forced European companies to halt AI deployments entirely, further widening the global digital divide.[4][5]