•
•
•
•
•
•
•
Transparency rules for AI content now apply in the EU, with a watermarking grace period ending on 2 December 2026. Here is who is affected and what they must do.
The European Union's Artificial Intelligence Act is the world's first comprehensive AI law. While much attention has focused on high-risk AI rules — which were postponed in 2026 — a different set of obligations is already in force: the transparency rules in Article 50.
These rules affect chatbots, image generators, video tools and anyone publishing deepfakes to EU audiences. Here is what they require.
This article provides general legal information, not legal advice. Businesses should consult qualified counsel about their specific obligations.
In May 2026, EU institutions agreed a Digital Omnibus package that changed several AI Act deadlines. According to analysis by Gibson Dunn:
| Obligation | Deadline |
|---|---|
| Article 50 transparency obligations | 2 August 2026 (unchanged) |
| Watermarking grace period for existing systems (Art. 50(2)) | Ends 2 December 2026 |
| High-risk AI systems (Annex III) | Postponed to 2 December 2027 |
| High-risk AI in regulated products (Annex I) | Postponed to 2 August 2028 |
In other words: high-risk rules were delayed, but transparency rules were not.
Providers must design AI systems that interact directly with people — such as chatbots and voice assistants — so users are informed they are dealing with an AI system, unless this is obvious from the context.
Providers of AI systems that generate synthetic audio, images, video or text must ensure outputs are marked in a machine-readable format and detectable as artificially generated. This is where watermarking and metadata standards come in.
Deployers of these systems must inform people exposed to them.
Deployers who create deepfakes — AI-generated or manipulated images, audio or video that resemble real people, places or events — must disclose that the content is artificial. Lighter rules apply to clearly artistic, satirical or fictional works.
Where AI-generated text is published to inform the public on matters of public interest, deployers must disclose that it was AI-generated — unless the content has undergone human review or editorial control and a person or organisation holds editorial responsibility. This is especially relevant for news publishers and content platforms.
The AI Act sets tiered fines. Breaches of obligations such as transparency duties can lead to fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher, with lower caps for SMEs.
Like the GDPR, the AI Act is likely to influence global standards. Indian companies serving EU users, content platforms and AI startups should expect clients and partners to ask about AI labelling and provenance.
Yes. They apply from 2 August 2026, with a grace period for machine-readable marking by existing systems until 2 December 2026.
Not necessarily. The text disclosure rule targets AI-generated text on matters of public interest, and there is an exception where human editorial review and responsibility apply.
They can be if they provide or deploy AI systems whose outputs are used in the EU.
No. They were postponed to December 2027 and August 2028.
Article 50 makes transparency the first AI Act obligation most businesses will actually feel. Clear disclosures, reliable content marking and strong editorial processes are now not just good practice — in the EU, they are the law.
Sources: EU Artificial Intelligence Act (Regulation (EU) 2024/1689); Gibson Dunn analysis of the AI Act Digital Omnibus agreement.
