The EU's Deepfake Labeling Code: What You Need to Know
The draft Code of Practice on AI-generated content marking introduces a new 'EU common icon' for deepfakes. Here's how it works and what providers must implement by August 2026.
Labeling AI-Generated Content
On December 17, 2025, the European Commission published the first draft of its Code of Practice on AI-generated content marking—a critical component of the AI Act's transparency obligations under Article 50.
What's Being Proposed
The draft Code introduces standardized methods for labeling AI-generated content, with the most notable proposal being an EU Common Icon (a universal symbol enabling people to identify at a glance whether content has been created or edited using AI), technical watermarking (machine-readable metadata embedded in AI-generated images, audio, and video), and disclosure requirements (clear textual disclosures for chatbots and interactive AI systems).
The EU Common Icon for Deepfakes
The proposed icon specifically targets synthetic media depicting apparently real events or persons. Key requirements: must be visible without requiring user action, cannot be easily removed or obscured, must appear in a consistent location across platforms, designed for accessibility (color-blind friendly, scalable).
Scope and Exceptions
The labeling requirements apply to AI-generated images depicting real persons or events, synthetic audio (voice cloning, AI-generated speech), AI-generated or manipulated video content, and text generated by conversational AI systems. Exceptions exist for artistic and satirical works (where obvious), authorized law enforcement activities, and content that's clearly fictional and doesn't depict real persons.
Technical Standards
The Code references C2PA (Coalition for Content Provenance and Authenticity) for cryptographic content credentials, IPTC metadata for standard fields for AI generation disclosure, and digital watermarking (invisible but detectable markers in media files).
Implementation Timeline
| Date | Milestone |
|---|---|
| December 2025 | First draft published for comment |
| March 2026 | Public consultation period ends |
| June 2026 | Final Code of Practice scheduled for publication |
| August 2, 2026 | Enforcement begins |
The Bigger Picture
This Code is the EU's answer to the global deepfake challenge. Implementation will require coordination between AI providers, platforms, and content verification services. Start planning your integration strategy now.