If your AI Act project plan has "high-risk compliance" somewhere in 2026, the Digital Omnibus just moved the goalposts. Good news: the deadline slid. Bad news: almost everything marketing actually touches didn't move an inch.
Here's the short version, then the details.
The Short Version
The Digital Omnibus on AI — Regulation (EU) 2026/1744 — amends the EU AI Act (Regulation (EU) 2024/1689). It was signed on 8 July 2026, published in the Official Journal on 24 July 2026, and entered into force on 27 July 2026.
It does two things marketers care about:
- It postpones the high-risk regime. The Chapter III (Sections 1–3) obligations for high-risk AI systems now start on 2 December 2027 for Annex III systems, and 2 August 2028 for high-risk AI embedded in regulated products (Annex I).
- It changes nothing else that marketing touches. Prohibitions, general-purpose AI (GPAI) rules, Article 50 transparency, AI literacy duties, penalties, and the 2 December 2026 machine-readable marking deadline for generative systems — all still apply, on the original dates.
The rules were delayed. They were not repealed.
What the Digital Omnibus Actually Changed
| Obligation | Before the Omnibus | After the Omnibus |
|---|---|---|
| High-risk systems under Annex III (biometrics, employment, credit scoring, education…) | 2 August 2026 | 2 December 2027 |
| High-risk AI in regulated products (machinery, medical devices, toys — Annex I) | 2 August 2027 | 2 August 2028 |
| Everything else in the AI Act | Original dates | Unchanged |
Three things worth knowing about the delay:
- It's a postponement of applicability, not a watering-down. The classification criteria, the conformity assessment route, and the obligations themselves are intact. You get more runway, not fewer duties.
- It was a competitiveness decision, not a safety finding. The Commission wanted to give industry breathing room while the GPAI rules and transparency duties bedded in. Expect scrutiny of how the extra time is used.
- Vendor timelines shift with it. If you buy tools that touch hiring, scoring, biometrics or emotion recognition, your suppliers' compliance evidence now lands in 2027 — which changes your procurement calendar, not your risk register.
What It Didn't Change — the Part Marketing Keeps Getting Wrong
This is where teams relax too early. The following are all live obligations today:
- Prohibited practices — in force since February 2025. That includes manipulative techniques, social scoring, untargeted scraping to build facial-recognition databases, and emotion recognition in the workplace. Marketing adjacent? Very.
- AI literacy (Article 4) — since February 2025, anyone deploying AI in their work needs a level of AI literacy. That's your whole marketing team, not just the data scientist.
- GPAI obligations — since August 2025, unchanged in substance and timing.
- Article 50 transparency — fully applicable since 2 August 2026. Chatbots must disclose they're AI. Synthetic content must be marked. Deepfakes must be labelled. Deployers — that's marketing — carry real disclosure duties. We keep a practical AI Act Article 50 checklist for marketing: chatbots, deepfakes and synthetic content for exactly this reason.
- The 2 December 2026 marking deadline — generative systems placed on the market before August 2026 must have machine-readable marking by then. If your content pipeline ingests AI-generated assets, that date is five minutes away, not 2027.
- Penalties — ceilings untouched. The Omnibus delayed the clock; it did not lower the fine.
Why Marketing Can't Relax
Because almost nothing marketing does is "high-risk" under Annex III — which means almost nothing marketing does was delayed.
Your team's exposure lives in the transparency and literacy layer that has been in force since August 2026 (and earlier). Every AI-written blog post, every generated image in a campaign, every chatbot on the pricing page — those duties are live now. The 2027 date changes the compliance calendar for HR-tech and biometric vendors; it changes nothing about the label you must put on a synthetic video this quarter.
The second reason is sneakier: your stack may contain someone else's high-risk system. Marketing automation increasingly includes scoring, targeting and profiling components. When those components are reclassified or re-certified in 2027, the evidence you'll need — DPIAs, conformity documentation, vendor attestations — has to exist in your files, not in a sales deck. Our security & compliance checklist for marketing automation is the fastest way to find out what you're already missing.
What to Do With the Extra Runway
Treat 2027 as a project date, not a holiday:
- Audit the stack now. List every AI feature in your martech and ask the vendor two questions: is any component high-risk or likely to be reclassified, and what evidence will exist by 2 December 2027?
- Fix the live duties first. Article 50 disclosures, synthetic-content marking, AI literacy training. These were due before the Omnibus and they're due after it.
- Put contract clauses in place at renewal. When tools come up for renewal in 2026–27, add AI Act representations: classification status, cooperation on conformity, change notification.
- Sequence the paperwork. DPIAs, vendor registers, incident procedures — the boring folder that turns a 2027 scramble into a formality.
The Realistic Timeline for Marketing Teams
| Date | What happens | Marketing action |
|---|---|---|
| Now (in force) | Prohibitions, AI literacy, GPAI, Article 50 transparency | Disclose, mark, train — already overdue if absent |
| 2 Dec 2026 | Machine-readable marking for generative systems placed on market before Aug 2026 | Verify marking survives your publishing pipeline |
| 2 Dec 2027 | High-risk obligations (Annex III) apply | Vendor evidence, contracts, DPIAs complete |
| 2 Aug 2028 | High-risk in regulated products (Annex I) apply | Only if you sell/embed in regulated products |
Practical guidance, not legal advice — we work alongside your legal counsel.