Three dates. Three different programs. One vote that made them official.
The European Parliament approved the AI Digital Omnibus on June 16, according to the parliamentary record, 423 in favor, 57 against, 174 abstentions. One formal step remains: Council adoption and Official Journal publication, expected before August 2, 2026. Member states’ ambassadors cleared the deal at Coreper on May 13, making that step procedural. The compliance calendar below reflects the post-Omnibus dates and takes legal effect upon OJ publication.
The practical consequence isn’t the vote. It’s that organizations running multiple AI systems, some generative, some use-based, some embedded in regulated products, now face three legally distinct obligation tiers with three different lead times. Treating them as one problem produces gaps. Treating them as three produces a program.
Tier 1: Article 50, December 2, 2026 (~5 months)
This is the near-term deadline. Article 50 transparency requirements apply to providers and deployers of generative AI and AI-generated content systems operating in the EU. Two obligations land simultaneously: mandatory standardized AI visual labeling on AI-generated text and media, and a new prohibition on nudifiers and CSAM introduced by the Omnibus. Organizations using generative AI to produce content for EU audiences are in scope. Five months isn’t a long runway when labeling infrastructure requires technical implementation and legal sign-off.
Who This Affects
Tier 2: Annex III, December 2, 2027 (~18 months)
High-risk system obligations for stand-alone Annex III systems, biometrics, critical infrastructure, HR, education, credit-scoring, apply 18 months out. That window looks comfortable until you count the steps: risk management system documentation, data governance framework, technical documentation package, human oversight architecture, and conformity assessment procedures. Most compliance teams underestimate the documentation lift.
The catch is that the scope of this tier isn’t fully settled. The European Commission’s draft Article 6 guidelines, published May 19, are in consultation until July 23, 2026. Law firm analysis of the draft, including Osborne Clarke’s assessment, suggests organizations shouldn’t assume their systems fall outside high-risk scope without conducting a documented conformity assessment review. Submit comments before July 23 and use the draft to run a preliminary scope assessment now, before the guidelines finalize.
Tier 3: Annex I, August 2, 2028 (~26 months)
AI embedded in regulated products under Annex I, machinery, medical devices, vehicles, faces the longest runway at 26 months. This tier’s compliance path runs through existing product safety regulation frameworks, which means the relevant documentation and conformity procedures are already partially governed by sector-specific law. Providers in this category should map their AI components against applicable product safety requirements before building a standalone AI Act program.
Next 90 Days: EU AI Act Action Items
- Monitor for Official Journal publication (confirms legal trigger)
- Run Article 50 scope assessment, identify all generative AI content touching EU audiences
- Submit to Article 6 consultation before July 23, 2026
- Conduct Annex III scope assessment using draft Commission guidelines
- Inventory Annex I product-embedded AI against applicable product safety regulations
One action item closes in four weeks. The Article 6 consultation deadline, July 23, 2026, is the only time-bounded action item in the next 30 days. Organizations with potential Annex III exposure should prioritize a scope assessment against the draft guidelines and decide whether to submit to the consultation before that window closes.
The real question is whether compliance programs built around a single August 2026 enforcement assumption have been reconfigured for a three-tier structure. Most haven’t. The organizations that reorient now, before the Article 6 guidelines finalize, will have a documented scope assessment before the deadline pressure builds.