Two Calendars, One Law: EU AI Act Transparency Is Live While High-Risk Rules Wait

August 2, 2026 was supposed to be the EU AI Act's high-risk reckoning. Instead, it became a tale of two calendars: transparency obligations went live while the heaviest conformity duties slipped to 2027 and 2028 after the bloc's AI Omnibus rewrote the timeline in late July.

For companies selling AI in Europe—and for U.S. labs with EU customers—the split matters more than any single fine headline. The law you must implement today is not the law everyone was briefing against six months ago.

What actually took effect on August 2

Goodwin's August 2026 client alert states plainly: Article 50 transparency and information obligations are now generally applicable and enforceable by national competent authorities across the EU.

These duties apply regardless of high-risk classification. If a product talks to users, generates synthetic images/audio/video/text, or scores emotions or biometrics, Article 50 likely applies—with a narrow exception for labeling legacy generative systems.

That is operational law as of this week, not a deferred milestone.

Examples of Article 50 scope cited in legal analyses include:

• Disclosing AI-generated content to users
• Marking deepfakes and synthetic media
• Informing users when they interact with an AI system (chatbots, voice agents)
• Transparency around emotion recognition and biometric categorization

Pharma commercial teams running patient-support chatbots, medical content generators, or recruitment AI in EU markets need Article 50 mappings now—even if their drug-discovery models won't face Annex III high-risk conformity until later.

What the Omnibus delayed—and to when

Regulation (EU) 2026/1744, the AI Omnibus, entered into force July 27, 2026, according to EUR-Lex consolidated text and Goodwin's summary.

It pushed high-risk obligations in Chapter III, Sections 1–3, to:

Disrupting Healthcare's July 2026 briefing for life-sciences teams notes the practical sequence: AI literacy and prohibited-practice controls should already exist; customer-facing transparency belongs in the 2026 plan; Annex III high-risk work shifts to late 2027; product-embedded medical AI to 2028.

The Omnibus rationale, per EUR-Lex recitals, cites delayed availability of harmonized standards, common specifications, and national competent authorities—implementation costs that would have spiked if high-risk rules had activated on the original August 2 date without supporting infrastructure.

Why pharma and frontier labs feel both timelines

Drug discovery AI rarely touches patients directly in early R&D—often argued as lower near-term regulatory risk. But the same companies deploy:

• Patient access chatbots (Article 50 now; potential Annex III if assessing eligibility for services)
• Clinical decision support (Annex III high-risk if influencing consequential decisions—deferred to 2027)
• AI-enabled devices (Annex I route—2028)

Merck's August 4 Exscientia and BenevolentAI deals highlight the business side: discovery alliances accelerate molecules years before regulatory classification bites. Legal teams must nonetheless track which deployed tools face August 2026 transparency duties versus 2027–2028 conformity packages.

Frontier AI labs face a parallel split. General-purpose models sold into the EU may trigger Article 50 provider obligations for synthetic outputs and user interaction disclosures even when the model itself is not (yet) listed as high-risk in a specific downstream use.

U.S. policy contrast in the same news cycle

While Brussels activated transparency and deferred high-risk conformity, Washington's AI policy conversation this week centered elsewhere: voluntary White House testing frameworks, Senate mandatory-access proposals, and UK AISI disclosures of unsanctioned agent behavior during cyber evaluations.

The transatlantic gap is structural. EU Article 50 creates immediate labeling and disclosure duties with Omnibus-adjusted dates for the heavier conformity chapters. U.S. federal AI governance remains a patchwork of agency action and pending legislation without an equivalent omnibus clock.

Companies operating in both jurisdictions cannot copy-paste a compliance calendar.

Operational checklist (analysis, not legal advice)

Legal counsel should validate any program. From public regulatory text and firm alerts, a sensible engineering-oriented checklist emerges:

1. Inventory EU-facing AI that generates content or interacts with natural persons.
2. Map Article 50 disclosures—mechanisms, languages, logging, update paths.
3. Separate deferred high-risk workstreams by Annex III vs Annex I product routes.
4. Track standardization outputs the Omnibus expects before 2027 conformity.
5. Do not treat Omnibus delay as standstill—transparency is enforceable now.

Bottom line

The EU AI Act's August 2, 2026 moment is real—but asymmetric. Transparency is live; high-risk conformity breathed until December 2027 and August 2028. Organizations that paused all compliance work when the Omnibus headlines landed may now be exposed on Article 50 while over-preparing for deadlines that moved.

The fix is dual-track governance: ship disclosure architecture this quarter; stage conformity evidence on the revised dates. Brussels gave industry more time for the heavy lift—and less excuse for ignoring the light one already in force.

### Sources

• Goodwin — Not Delayed, Not Deferred: EU AI Act Transparency Obligations Are Now in Force (August 2026)
• EUR-Lex — Regulation (EU) 2026/1744 (AI Omnibus) (July 27, 2026)
• EUR-Lex — Consolidated AI Act text as of July 27, 2026 (July 2026)
• Disrupting Healthcare — EU AI Act Deadlines for Life Sciences: 2026–2028 (July 2026)