EU AI Comply
EN
Start free
← Blog 3 min

What stays mandatory on 2 August 2026 after the Omnibus — Article 50, in plain language

If you've read recently that "the AI Act got postponed", you've read half a sentence. The other half — the one that matters for most companies — is that the transparency obligations in Article 50 apply exactly as scheduled: from 2 August 2026. Let's sort this out calmly, without jargon.

What actually got postponed

The Digital Omnibus on AI was adopted by the European Parliament on 16 June 2026, given final approval by the Council on 29 June, and signed on 8 July. Publication in the EU's Official Journal is next; the act enters into force three days after publication. Its most consequential change:

  • Stand-alone high-risk AI systems (recruitment, credit scoring, education, biometrics, critical infrastructure — Annex III): obligations move from 2 August 2026 to 2 December 2027.
  • High-risk AI embedded in regulated products (medical devices, machinery, toys — Annex I): the deadline becomes 2 August 2028.

If your company was preparing for high-risk compliance, you got time. That is what was postponed — and only that.

What was NOT postponed

Article 50 — the transparency obligations — applies from 2 August 2026. In short, it says four things:

  1. If a person is talking to an AI, they must know it. Chatbots and assistants interacting with the public must clearly disclose they are AI — unless it's obvious to any reasonable person.
  2. AI-generated content must be marked (synthetic audio, images, video, text) in a machine-readable format. Here sits the one nuance the Omnibus added: systems already on the market before 2 August 2026 get until 2 December 2026 for the marking obligation. Systems launched after 2 August must comply from day one.
  3. Emotion recognition and biometric categorisation must be disclosed to the people exposed to them.
  4. Deepfakes and AI-generated text on matters of public interest must be labelled as such — unless a human carries editorial responsibility.

And the penalties for breaching these obligations are unchanged: up to €15 million or 3% of global turnover, whichever is higher.

The five most common misunderstandings

1. "The whole AI Act was postponed." No. The high-risk obligations were. Transparency (Art. 50), the prohibitions (enforced since February 2025), the rules for general-purpose AI models (since August 2025) and the staff AI-literacy duty all remain in force.

2. "We're a small company, this isn't about us." Article 50 has no size threshold. The Omnibus SME simplifications target other chapters (quality-management systems, high-risk documentation) — not transparency.

3. "We only use AI, we don't build it." Professional users (deployers) have obligations too: if you publish generated content, run a customer-facing chatbot, or use emotion-analysis tools, parts of Article 50 apply to you directly.

4. "Our country hasn't even designated an authority yet, so it doesn't apply." An EU regulation applies directly, whatever the state of national legislation. France and the Netherlands, for instance, have not yet formally designated their authorities — but obligations for companies there still start on 2 August 2026.

5. "Let's wait for the Official Journal, maybe something changes." The act has been signed since 8 July; publication is a formality expected before 2 August. No version of the text moves the Article 50 deadline.

What you can do this week

  1. Take inventory: which AI systems your company uses or offers — including "just" conversational assistants.
  2. Find your Article 50 touchpoints: do you run a public chatbot? Publish generated content? Use emotion analysis? Produce or distribute deepfakes (including in marketing)?
  3. For new systems, plan content marking from launch — there is no grace period for them.
  4. Document what you checked and what you decided — evidence of diligence matters in front of any authority.
  5. Don't forget AI literacy: the Omnibus softened the duty ("support the development of" skills rather than "ensure" them), but it did not remove it.

That is exactly what our platform automates: inventory, risk classification and an obligation plan with deadlines — including the Article 50 transparency workflow.

Sources