AI · Regulation
The EU AI Act was delayed, but not all of it: what changes on December 2
The Digital Omnibus pushed high-risk rules to 2027, but transparency duties already apply and another deadline lands on December 2. Why it matters to a Dominican company selling into Europe.
If your company builds a chatbot, generates images or text with AI, or integrates language models into a product used by people in Europe, this matters to you even if you operate from Santo Domingo. The EU Artificial Intelligence Act changed its timeline this year, and many people read “delayed” and stopped paying attention. That’s a mistake.
What was delayed
The European Union adopted the Digital Omnibus on AI, which entered into force on July 27, 2026 and moved the law’s heaviest obligations:
- Stand-alone high-risk systems —those used for hiring, credit scoring, exam grading, policing or border control— were due to comply on August 2, 2026. The new date is December 2, 2027.
- AI embedded in regulated products (medical devices, toys, lifts) moves to August 2, 2028.
The reason was practical: the technical standards companies need in order to comply weren’t ready.
What already applies (since August 2)
The Article 50 transparency rules did not move. Since August 2, 2026:
- If your system talks to people, it must tell them they’re interacting with AI, unless that’s obvious.
- If your system generates text, images, audio or video, those outputs must be marked in a machine-readable format so they can be detected as AI-generated.
- Anyone who publishes deepfakes, or AI-generated text meant to inform the public on matters of public interest, must label them as such.
What lands on December 2, 2026
Two things, seven weeks from now:
- Marking generated content. Generative AI systems already on the market before August 2 got four extra months only for machine-readable marking. That window closes on December 2. Systems launched after August 2 had to comply from day one.
- New bans. From that date it is illegal to offer in the EU systems designed to generate child sexual abuse material or non-consensual intimate imagery of real people.
Maximum fines are unchanged: up to €35 million or 7% of global turnover for prohibited practices, and up to €15 million or 3% for most other breaches.
Why does it matter to a company in the DR?
Because the law doesn’t look at where you are, but at where your system is used. If your product reaches users in Europe —a client in Spain, an app for the diaspora, a SaaS with customers in the Netherlands— you fall within its scope as a provider or as a deployer.
A short checklist for this month
- Inventory. Which features of your product use AI, and which reach users in Europe?
- Notices. Does your chatbot clearly say it’s an AI?
- Marking. Does what your system generates carry metadata or marks that allow detection? In June the Commission published a code of practice calling for layered measures: metadata, watermarks and detection tools.
- Risk. If your product touches hiring, credit or education, start documenting data, risks and human oversight now. December 2027 will come fast.
How we handle it
At Hekonix we design AI features with transparency built in from the start. In MedicoNet, for example, everything the AI writes is flagged for review and nothing is saved without the doctor’s confirmation. That same principle —people know when AI is involved and keep the final word— is what the regulation is asking for.
If your product uses AI and has users abroad, let’s talk.
This article is for information only and is not legal advice. For your specific case, consult a lawyer who specializes in EU regulation.