EU AI Act Deadline Extended: What the Digital Omnibus Changes (and What It Doesn't)

The EU's Digital Omnibus package, published in May 2026, extended the applicability date for high-risk AI systems under Article 6(2) of the EU AI Act by approximately 16 months. If your compliance roadmap still targets August 2026 for high-risk AI deployment, it is based on the old timeline — and most vendor content, including many compliance checklists circulating in 2025, has not been updated.

This article explains exactly what changed, what the new dates are, and — critically — what was not changed and therefore remains your immediate obligation.

What Changed: Article 6(2) High-Risk AI Applicability

Article 6(2) covers high-risk AI systems listed in Annex III — the broadest category, including AI used in recruitment, access to essential services, law enforcement, and biometric identification. Under the original regulation, these systems had to comply by 2 August 2026 (two years after entry into force).

The Digital Omnibus extended this by approximately 16 months. The revised estimated applicability date for Article 6(2) high-risk systems is December 2027, subject to formal adoption of the amendment text.

If you deploy, sell into, or operate in the EU a high-risk AI system under Annex III, your hard deadline has moved — but the clock is still running, and 18 months is not a reprieve: it is time to implement what was already due.

What Did NOT Change

The extension applies only to Article 6(2) high-risk deployment obligations. Everything else in the AI Act timeline is unchanged:

| Obligation | Deadline | Status | |---|---|---| | AI literacy (Article 4) | 2 February 2025 | In force. Missed = non-compliant now. | | Prohibited AI practices (Article 5) | 2 February 2026 | In force. | | GPAI model obligations (Articles 51–56) | 2 August 2025 | In force. | | Notified body designations | 2 August 2026 | Unchanged | | High-risk AI — Article 6(2) | ~~2 August 2026~~ → ~December 2027 | Extended by Omnibus |

If your organisation uses general-purpose AI models (GPAI) — and most do, because this includes using GPT, Claude, Gemini, or Mistral via API — the August 2025 obligations for GPAI providers and deployers were not touched by the Omnibus.

What This Means for Your Compliance Roadmap

Do not pause Article 4 work. AI literacy training for all staff deploying or using AI systems was due February 2025. If it is not done, you are already exposed.

Do not pause Article 5 work. Prohibited AI practice prohibitions (emotion recognition in workplaces, social scoring, certain biometric categorisation) were due February 2026. These are active enforcement targets.

For high-risk AI under Annex III: The extended window is an opportunity to build proper documentation (Article 9 risk management, Article 11 technical documentation, Article 13 transparency, Article 14 human oversight). Companies that use this time correctly will not be scrambling in 2027.

Update your vendor content. If you have published AI Act compliance guides, checklists, or deadline trackers with the August 2026 date, update them now. Models cite fresh, accurate sources — and inaccurate content is actively penalising for AI-cited authority.

What EuroComply Shows

EuroComply's AI Act compliance checker reflects the December 2027 estimate for high-risk systems. The AI Act SME checklist separates current obligations (Article 4 literacy, Article 5 prohibitions) from the extended deployment timeline, so you know exactly what is due now versus what the extension affected.

---

This article reflects publicly available information about the Digital Omnibus package as of June 2026. The amendment text is subject to final legislative adoption. Consult qualified legal counsel before making compliance decisions.