EU transparency duties are live while high-risk rules wait: the two-track reality

EUMAJORStatute in forceContent labelingChatbots & disclosurePenalties

What changed

Since August 2, 2026, chatbots must disclose they are AI, synthetic content must be machine-readably marked (grace until December 2, 2026), and fines up to 3% of global turnover are live — while the headline high-risk obligations moved to December 2027.

Who this touches

Any organization with customer-facing AI serving EU users, and anyone generating synthetic content for EU audiences.

Why it matters

The headline everyone will repeat is that Europe delayed. The part that matters is what didn't move: the transparency layer, the GPAI enforcement machinery, and the fines.

Suggested next step — and why

Inventory every customer-facing AI touchpoint for disclosure coverage, and stand up content marking before December 2 — one build can satisfy the EU, China, Korea, and California rules at once.

The clock

Marking grace ends December 2, 2026. High-risk obligations: December 2, 2027.

The editor's read

Everyone will cover the deferral. The story worth your attention is the marking deadline that didn't move — December 2 is the first date on which an ordinary company can be fined for something its marketing team did. — IDV

Source: Official Journal of the EU — primary text — read in the original, not from coverage.

Permanent record: this page is public forever. "When did we know?" always has a URL.

Curated nightly by Dr. Ivan Del Valle — AI governance architect, editor of The Compact. Disagree with a reading? Write ivan.delvalle@rsit.ai — every message is read, most are answered. Corrections are published, not buried. · Today's brief · Method · Reporting free forever; the desk is sustained by its data plans.