Skip to content

Regulation

AI content must be disclosed.
Real content must be provable.

Article 50 of the EU AI Act has applied since 2 August 2026. Here is what it requires, what changed with the Digital Omnibus, and where physical-layer authentication fits. Last reviewed 2 November 2026. This page is general information, not legal advice.

Key dates

  1. 2 Aug 2026

    Article 50 applies

    Transparency obligations become applicable and enforceable by national authorities, including deepfake disclosure by deployers.

  2. 2 Dec 2026

    Marking grace period ends

    Under the Digital Omnibus on AI (Regulation (EU) 2026/1744), generative systems already on the market before 2 August 2026 have until this date to meet the machine-readable marking duty in Article 50(2).

  3. Dec 2027 / Aug 2028

    High-risk obligations

    The Omnibus postponed most high-risk system obligations. It did not postpone Article 50.

Article 50 in brief

Four transparency obligations

AI that interacts with people

Providers must design systems so people are informed they are interacting with AI, unless obvious from context.

Synthetic content marking

Providers of generative AI must mark audio, image, video and text outputs in a machine-readable format so they are detectable as artificially generated or manipulated.

Emotion recognition & biometric categorisation

Deployers must inform the people exposed to these systems.

Deepfakes & AI-generated public-interest text

Deployers must disclose that deepfake content, or AI-generated text published to inform the public, has been artificially generated or manipulated, subject to defined exceptions.

Fines for breaches of Article 50: up to €15 million or 3% of worldwide annual turnover. The European Commission published guidelines on the transparency obligations in July 2026.

The half the regulation can't cover.

Article 50 makes honest AI providers label their output. It cannot make bad actors do the same, and labels can be removed.

That leaves broadcasters, institutions and platforms with an open question every time a clip goes viral: is this real?

Lightmark answers it from the other side. Footage captured in a Lightmark-signed space carries a dynamic signature in its pixels. Content that was never there, including unlabelled synthetic content, does not.

Questions, answered.

Did the Digital Omnibus delay Article 50?

No. The Digital Omnibus on AI delayed most high-risk obligations to December 2027 and August 2028, but Article 50 transparency obligations applied from 2 August 2026. The only adjustment is a grace period to 2 December 2026 for the Article 50(2) marking duty for generative systems placed on the market before 2 August 2026.

What are the penalties?

Breaches of Article 50 can lead to administrative fines of up to €15 million or 3% of total worldwide annual turnover, whichever is higher.

Does Lightmark make an organisation compliant with Article 50?

No single tool does. Article 50 places duties on providers and deployers of AI systems. Lightmark addresses the complementary problem: proving that authentic footage is authentic, so genuine content can be distinguished from unmarked or relabelled synthetic content.

Why isn't marking AI content enough?

Marks and labels can be stripped by re-encoding, cropping or screen recording, and bad actors will not mark their output. Physical-layer evidence lets authentic footage prove its own origin regardless of what happens to fakes.

Make your authentic footage provable.

Talk to us about where Lightmark fits alongside your AI Act programme.

Book a briefing