The EU’s AI transparency rules just kicked in.

For two years, 2 August 2026 sat in compliance calendars as the day the EU AI Act would switch on in earnest. The date arrived — but not the way most of last year’s articles predicted. A late amendment, the Digital Omnibus on AI signed on 8 July 2026, postponed the heavyweight high-risk obligations to late 2027 and 2028. What survived intact, and now applies, is Article 50: the transparency rules. And those are precisely the ones that touch marketing.

What applies since 2 August

Three duties, in plain terms. First, AI systems that interact with people — chatbots, voice assistants, AI agents — must make clear the person is talking to a machine, unless that is obvious from context. Second, deepfakes and AI-manipulated audio, image, or video must be disclosed as artificially generated, and AI-generated text on matters of public interest must be disclosed as well. Third, emotion-recognition and biometric-categorisation systems must inform the people exposed to them. The European Commission adopted interpretive guidelines on 20 July 2026, and the penalty ceiling is real: up to €15 million or 3% of worldwide annual turnover.

Why this reaches ordinary businesses

The transparency rules are the most widely applicable part of the whole Act, because they bind not only providers of AI systems but also deployers — companies that merely use them. A brand with a branded chatbot on its website counts. A business publishing synthetic video in its campaigns counts. This is not a Big Tech regulation with a trickle-down effect; it applies directly to the marketing department.

What it does not require

Two reliefs worth knowing. Content generated and published before 2 August 2026 does not need retroactive labels. And the machine-readable marking duty — watermarking synthetic output so it is detectable as AI-generated — carries a grace period to 2 December 2026 for systems already on the market; in practice, most businesses will rely on what their model providers ship to satisfy it. The everyday use of AI as a tool — drafting, editing, research assistance with human review — is not what Article 50 targets; the duties bite where people interact with a machine or consume synthetic media without knowing it.

What marketers should do now

Inventory first: list every place your brand deploys AI that customers meet — chatbots, voice agents, synthetic video or audio in ads, AI-generated imagery of real-seeming people. Label the interactions and the synthetic media clearly. Ask your AI vendors what marking they provide. And treat the wave of outdated ‘AI Act 2026’ articles with caution — much of what was published before July describes obligations that no longer arrive this year. We are marketers, not lawyers; for edge cases, confirm with counsel.

The strategic read

Disclosure is becoming the norm at exactly the moment synthetic content floods every channel — and that plays to brands that lean on verifiable, human-made substance. The label ‘made with AI’ costs nothing to those with nothing to hide, and quietly re-prices everyone else. If your positioning is built on authenticity, this regulation is tailwind, not burden. That is how we read it at FIB — and how we build for clients: get in touch.

Sources

European Commission, guidelines on Article 50, 20 July 2026 · Regulation (EU) 2024/1689 (AI Act) · Digital Omnibus on AI, 8 July 2026 · Cooley LLP, August 2026 · Stibbe, July 2026 · Norton Rose Fulbright Data Protection Report, July 2026 · Technology.org, July 2026.