EU AI Act Article 50 Deepfake Disclosure Guide 2026

The countdown has started. On August 2, 2026, the EU AI Act enters full effect — and Article 50 introduces the strictest deepfake disclosure requirements the world has seen. For any brand, agency, media company, or content platform reaching European audiences, this is not an optional compliance exercise. Penalties reach €15 million or 3% of global annual turnover.

The good news: there is a clean, practical path to compliance that avoids disclosure requirements entirely — and BGBlur is central to that workflow.

This guide breaks down exactly what Article 50 requires, who is in scope, how the law compares to US regulations, and how face blur on real human footage provides a disclosure-free alternative to AI-generated synthetic content.

What Is EU AI Act Article 50?

The EU Artificial Intelligence Act (Regulation (EU) 2024/1689) is the world's first comprehensive legal framework for AI regulation. It categorizes AI systems by risk level and imposes obligations proportionate to that risk.

Article 50 specifically addresses transparency requirements for AI-generated content. It targets content that could mislead people into believing they are watching real events or real people doing or saying things they did not actually do or say.

The Core Disclosure Obligation

Article 50(4) states that providers of AI systems that generate or manipulate image, audio, or video content constituting a deepfake must ensure that the output is marked in a machine-readable format and is detectable as artificially generated or manipulated. Where technically feasible, the content must carry a visible disclosure that is clearly perceptible to the audience.

In practice, this means:

• AI-generated faces in video must be labeled
• Synthetic voices deployed on real persons' likenesses must be labeled
• AI-manipulated footage that creates a false impression of real events must be labeled
• The label must be embedded in the content (machine-readable) AND visible to viewers

What Counts as a "Deepfake" Under Article 50?

Article 50 defines the scope broadly. Content falls under the disclosure requirement when it:

1. Depicts real persons — public figures or private individuals — in scenarios they were not actually in
2. Uses AI to generate or significantly manipulate the audio or visual representation of a real person
3. Creates a realistic impression that the person said or did something they did not

This includes:

• AI-generated video using someone's likeness
• Voice cloning deployed on video of real people
• Face-swap deepfakes
• AI-enhanced footage that materially alters what a person said or did
• Synthetic performers created to resemble or stand in for real individuals

What Is Exempt?

Article 50 includes an exemption for satire and parody — but with a critical condition: the content must still be clearly labeled as artificially generated. The exemption removes the requirement that disclosure be "clearly perceptible" (i.e., front-and-center visible) but does not remove the machine-readable labeling requirement entirely.

Crucially, advertising and commercial content does not qualify for the satire exemption, regardless of how humorous the intent.

Who Is in Scope?

Article 50 applies to:

Providers — companies or individuals that develop AI systems used to generate or manipulate content

Deployers — companies or individuals that use those AI systems to produce content for distribution

Scope is geographic, not entity-based. A US company producing AI-generated video that is accessible to EU residents is in scope. A Brazilian influencer posting deepfake content on a platform with EU users is in scope.

In practical terms, this affects:

• Advertising agencies using AI-generated models or synthetic spokespersons
• Media and news organizations using AI to reconstruct or enhance footage
• Social media creators deploying face-swap or voice-clone effects
• Marketing teams using virtual influencers or AI-generated brand ambassadors
• Entertainment companies using AI de-aging, face replacement, or posthumous content generation
• Political campaigns using AI-generated campaign materials depicting candidates

The Penalty Structure

Violations of Article 50 transparency obligations fall under the EU AI Act's tiered penalty structure:

For a mid-size company with €500M annual global revenue, a 3% penalty represents a €15M fine — and that is on a per-violation basis. A single advertising campaign with undisclosed deepfake content could constitute multiple violations.

National data protection authorities (the same bodies that enforce GDPR) will generally serve as the competent authority for Article 50 enforcement.

Article 50 vs NY S.8420-A: The Transatlantic Compliance Problem

Brands operating in both the US and EU now face overlapping but non-identical disclosure requirements.

The key insight from this comparison: both laws exempt real human content. Only synthetic or AI-manipulated depictions trigger mandatory disclosure. This makes real-human workflows with appropriate privacy protections (like face blur) the universal compliance solution for both jurisdictions.

For a detailed breakdown of the NY S.8420-A requirements, see our guide to the New York AI disclosure law for synthetic performers.

The BGBlur Compliance Strategy: Real Humans + Face Blur

Here is the central compliance insight:

Face blur on real human footage is not AI-generated content under Article 50.

When you shoot real people and apply face blur:

• The performers are real humans who actually performed the actions shown
• Face blur protects their identity but does not fabricate their actions or statements
• No AI is generating a synthetic likeness or voice
• The content does not create a false impression of what a real person said or did

Article 50 does not apply. No disclosure is required.

This is in sharp contrast to:

• Generating a synthetic face using Midjourney, DALL-E, or similar tools — disclosure required
• Using an AI voice clone on real footage — disclosure required
• Face-swapping a celebrity onto an actor's performance — disclosure required
• Creating a virtual influencer using generative AI — disclosure required

The BGBlur Workflow for Article 50 Compliance

Step 1: Shoot with real human performers

Use real actors, employees, or talent for your video content. This is the foundational step — the content must depict genuine human activity.

Step 2: Apply face blur as needed

Upload the footage to BGBlur. Apply face blur to protect performer identity (if required), protect bystanders, or as a creative/brand style choice. BGBlur's AI detects and tracks faces automatically across all frames.

Step 3: Export and publish

The exported video shows real human activity. No AI-generated or AI-manipulated synthetic depiction is present. Article 50 disclosure is not triggered.

Step 4: Document your workflow

Maintain records showing that your content features real human performers. Under EU compliance frameworks, documentation is key — you want to be able to demonstrate, if audited, that the content does not involve AI-generated synthetic depictions.

Article 50 Compliance Checklist

Use this checklist before publishing any video content to EU-accessible platforms after August 2, 2026:

Content audit:

• Does the video use any AI-generated faces or synthetic performers?
• Does the video use voice cloning or AI-generated audio attributed to a real person?
• Does the video depict a real person doing or saying something they did not actually do or say?
• Was any generative AI used to materially alter or reconstruct visual footage of a real person?

If any box above is checked:

• Add machine-readable AI content label (C2PA standard or equivalent)
• Add viewer-visible disclosure in the content itself
• Confirm the disclosure is perceptible at normal viewing speed (not buried in credits)
• Document the AI systems used and how disclosure was implemented

If no boxes are checked (real human content):

• Confirm all performers are real human beings
• Apply face blur or other identity protection as required by your data protection obligations
• Document that no AI generation or significant manipulation of human likeness was used
• No Article 50 disclosure required

For broader data protection compliance for your video content, also review our GDPR video compliance guide and, if publishing in the UK, the UK GDPR video privacy guide.

Practical Scenarios: Disclosure Required or Not?

Scenario 1: Fashion brand ad featuring AI-generated models AI-generated synthetic faces replace real models. Result: Article 50 disclosure required.

Scenario 2: Corporate training video with real employees, faces blurred for privacy Real human performers, face blur applied. Result: No disclosure required under Article 50.

Scenario 3: News package using AI to reconstruct damaged archival footage of a historical figure AI manipulation of a real person's likeness. Result: Article 50 disclosure required.

Scenario 4: Social media campaign using real talent, with face blur for brand consistency Real human performers, face blur applied as a stylistic or privacy choice. Result: No disclosure required under Article 50.

Scenario 5: Virtual influencer created with generative AI promoting a product Fully synthetic performer. Result: Article 50 disclosure required (and potentially NY S.8420-A as well if targeting New York).

Scenario 6: Testimonial video with real customer, face blurred at their request Real human, face blur for identity protection. Result: No disclosure required under Article 50. (Also consider voice anonymization if the customer's voice is also identifiable.)

What Happens After August 2, 2026?

Enforcement will ramp up progressively. The European AI Office — established under the AI Act — coordinates enforcement across member states. National regulators (the same bodies that have levied significant GDPR fines against Meta, Google, and TikTok) serve as local enforcers.

Based on GDPR enforcement patterns, initial enforcement actions will likely focus on:

• High-profile advertisers with clearly non-compliant content
• Political advertising ahead of elections
• Consumer-facing platforms with large volumes of user-generated deepfake content

The window to build compliant workflows is now. Brands that establish real-human-with-blur workflows before August 2026 will be in a substantially stronger compliance position than those scrambling to retrofit disclosure labels after enforcement begins.

Summary

• EU AI Act Article 50 takes effect August 2, 2026 and mandates disclosure of AI-generated and AI-manipulated video content
• Scope covers deepfakes, synthetic faces, voice cloning on real persons, and AI-manipulated footage creating false impressions
• Penalties reach €15M or 3% of global annual turnover — applicable to any entity whose content reaches EU users
• Real human footage with face blur does not trigger Article 50 disclosure — the content is not AI-generated
• BGBlur provides the browser-based face blur workflow that keeps real-human content compliant with both EU AI Act Article 50 and US regulations like NY S.8420-A
• Start building compliant workflows now — enforcement begins August 2026