EU AI Act: postponed high-risk deadlines to 2027-2028 reshape industry timelines

On 7 May 2026, the EU Council and Parliament agreed an 'omnibus package' postponing the AI Act's heaviest deadlines by twelve to eighteen months. Standalone high-risk AI systems under Annex III (HR, credit scoring, border control, education) now apply from 2 December 2027 instead of 2 August 2026. AI systems embedded in already-regulated products — industrial machinery, medical devices, vehicles, toys — shift from 2 August 2027 to 2 August 2028. Member states now have until August 2027 to stand up their regulatory sandboxes.

Cardan-AI's takeaway: this delay removes no obligation — it resets the preparation clock. For aerospace and O&G, where AI increasingly sits inside Annex I machinery and equipment (predictive maintenance, automated inspection, safety systems), the window now extends to mid-2028 — valuable time to build robust compliance documentation rather than rush it under deadline pressure. But the Article 50 transparency obligation (disclosing AI interaction to users) still applies from 2 August 2026, with only a four-month grace period on watermarking.

The classic blind spot: many industrial players will read 'postponed' as 'we have time' and shelve compliance work until 2027. That mirrors the opposite mistake of 2025's rush — a regulatory calendar shift is not a shift in technical due diligence. Mapping AI systems against the AI Act's annexes and documenting datasets and risk use-cases remains months of groundwork, done better without deadline pressure.

Cardan-AI recommendation: use this window to turn a regulatory constraint into a competitive edge — a documented, audited system reassures clients and partners well ahead of the legal deadline. Immediate priority regardless: verify your conversational interfaces and generative tools already meet Article 50 transparency, the one obligation still locked to 2 August 2026.