The EU AI Act deadline everyone was bracing for just got rewritten — six days before it hit. The Digital Omnibus entered into force on July 27, 2026, five days ahead of the AI Act’s original August 2, 2026 high-risk compliance date. If you’d been planning around that date, some of what you were preparing for just moved. Some of it didn’t. Here’s what actually changed, what’s still enforceable right now, and what to do about both.
What the Digital Omnibus Changed About the EU AI Act Deadline
Regulation (EU) 2026/1744, known as the Digital Omnibus on AI, was published in the Official Journal on July 24, 2026 and took effect three days later. It’s worth being precise about what it did: it deferred the high-risk compliance regime, not the AI Act as a whole.
Two new dates replace the original August 2, 2026 target. Standalone high-risk AI systems under Annex III — hiring and recruitment tools, credit-scoring and lending engines, biometric identification, systems used in education, and AI supporting essential public services — now have until December 2, 2027 to meet the full conformity-assessment regime. High-risk AI embedded in products already covered by EU product-safety law (Annex I) has until August 2, 2028.
What’s Still Due on August 2, 2026
The deferral only covers high-risk systems. Three obligation sets stayed on the original date, and they’re enforceable now, not on some future timeline:
- Article 50 transparency duties. If your product includes a chatbot or conversational AI feature, it has to disclose that people are talking to an AI. If it generates or lets users generate synthetic images, audio, video, or text, that content has to be labeled as AI-generated. Emotion-recognition and biometric-categorization features have to disclose their use to the people they’re applied to.
- GPAI provider obligations. Rules for providers of general-purpose AI models have been in force since August 2, 2025, and the Omnibus didn’t touch them.
- The Article 5 prohibited-practices regime. In force since February 2, 2025 — also unaffected.
There’s one narrow exception worth knowing inside Article 50 itself: the sub-duty covering AI-generated content marking and detection (Article 50(2)) got its own short grace period, extended to December 2, 2026 — but only for systems that were already on the market before August 2, 2026. Every other Article 50 obligation, including the basic chatbot-disclosure duty, applies today.
Does This Apply to Your Product?
Scope isn’t limited to EU-headquartered companies. A US or South African vendor selling into the EU market, or whose product reaches EU-based users, is in scope the same as a company based in the EU.
Run a quick check against your own product:
- Does it include a chatbot, virtual assistant, or any conversational AI feature? Article 50 disclosure applies to you now.
- Do you generate, or let users generate, synthetic images, audio, video, or text? Labeling duties apply now — and the narrower marking/detection sub-duty may qualify for the grace period if you were already on the market before August 2026.
- Does your product touch hiring, lending or credit decisions, biometric identification, education assessment, or essential public services? You’re likely looking at Annex III high-risk classification — now on the December 2027 runway instead of this year.
“High-risk” and “requires an AI disclosure” aren’t the same question. A product can trigger Article 50 transparency duties today without being anywhere near the Annex III high-risk category, and most companies building anything with a visible AI interface need to check both.
What Happens If You Miss These
Article 99 sets three penalty tiers. Prohibited practices carry fines up to €35 million or 7% of global annual turnover. High-risk and other obligations, including Article 50 transparency, top out at €15 million or 3% of turnover. Supplying misleading information to authorities caps at €7.5 million or 1% of turnover. The Article 50 transparency duties aren’t a 2027 problem waiting to happen — they’re enforceable today, at the €15 million/3% tier.
FAQ
What is the current EU AI Act deadline?
It depends which obligation you mean. Article 50 transparency duties, GPAI provider rules, and the prohibited-practices regime are all already in force, unaffected by the recent change. High-risk Annex III compliance now runs to December 2, 2027, and Annex I to August 2, 2028, following the Digital Omnibus deferral.
Did the EU delay the AI Act?
Partially. The Digital Omnibus (Regulation (EU) 2026/1744, in force July 27, 2026) delayed the high-risk compliance regime specifically. Transparency, GPAI, and prohibited-practices obligations were left on their original schedule.
Does the EU AI Act apply to companies outside the EU?
Yes. Scope follows the market, not the company’s headquarters — a non-EU vendor whose product reaches EU-based users or is sold into the EU market is in scope.
What are the penalties for non-compliance with the EU AI Act?
Up to €35 million or 7% of global turnover for prohibited practices, up to €15 million or 3% for high-risk and other obligations (including Article 50), and up to €7.5 million or 1% for misleading information supplied to authorities.
Where This Leaves You
The high-risk deferral is real breathing room, not a reprieve from the whole regulation. If your product has any AI-facing feature — a chatbot, generative content, biometric or emotion-recognition functionality — the transparency work is due now, not in 2027. And the extra runway on Annex III and Annex I obligations is best used building a real compliance program, not sitting on it until the next deadline sneaks up the same way this one did.
We approach the EU AI Act deadline the way we approach any engineering problem: as product and process changes, not a form to fill out or a scanner to bolt on. If you’re still working out which of your features trigger Article 50 today, or you know you’re heading toward Annex III high-risk classification and want the runway used well, talk to us about what an EU AI Act compliance partner can do for your team.
Image credits: EU flag photo by Engin Akyurt on Unsplash. AI chatbot photo by Igor Omilaev on Unsplash.
