Article 50 transparency obligations apply from 2 August 2026, but the duties vary by provider and deployer role and by use case. This tool checks your pipeline against the binding Article 50 baseline plus the current second-draft code best practices.

Scope: Article 50 covers chatbots, deepfake/synthetic media generators, emotion recognition systems, and AI text generators on public interest matters. Under current law, the baseline deadline remains 2 August 2026. Proposal-stage Omnibus talks have not changed that date.

Check Your Content Pipeline

Answer six questions mapping to the current second-draft code's two-layer marking approach, plus practical labelling and detection controls that help teams operationalise Article 50.

Privacy By Design: This executes entirely in your browser. We never see your responses.

1. Content Type Identification

Which AI-generated content types does your organization produce or deploy?

Chatbot / Conversational AI

Users interact with an AI system (customer service bots, virtual assistants, AI copilots). Article 50(1) applies.

Synthetic Media (Images, Audio, Video)

AI generates or manipulates visual/audio content (marketing images, voice synthesis, video editing). Article 50(4) applies.

AI-Generated Text on Public Interest

AI produces text published on matters of public interest (news, analysis, reports). Article 50(4) applies.

Emotion Recognition / Biometric Categorization

AI infers emotional states or categorizes biometric attributes. Article 50(3) applies.

None of the Above

Backend-only AI with no public-facing content generation or user interaction.

2. Layer 1: Secured Metadata

Does your content pipeline embed machine-readable provenance metadata (e.g., C2PA content credentials) in generated files?

No Metadata (Gap)

Generated content contains no provenance information. Origin cannot be traced programmatically.

Basic Metadata

Some metadata exists (EXIF, basic tags) but not in a standardized provenance format like C2PA.

C2PA or Equivalent (Compliant)

Content credentials are embedded using C2PA or equivalent standard. Provenance is machine-readable and tamper-evident.

3. Layer 2: Watermarking

Does your pipeline embed imperceptible watermarks in AI-generated content that can be detected by automated tools?

No Watermarking (Gap)

Content is generated without any embedded watermark signal.

Visible Watermark Only

A visible label or overlay exists but no imperceptible machine-detectable watermark is embedded in the content data.

Imperceptible Watermark (Compliant)

Content embeds imperceptible watermarks that survive common transformations (compression, cropping) and can be detected by automated verification tools.

4. User-Facing Labelling

Do you provide clear, visible labels to end users indicating content is AI-generated or that they are interacting with an AI system?

No Labelling (Gap)

Users are not informed that content is AI-generated or that they are interacting with AI.

Buried Disclosure

Disclosure exists in terms of service or footer text but is not prominent at the point of interaction or consumption.

Prominent Labelling (Compliant)

Clear, visible labelling at the point of interaction/consumption. Users are informed before or during engagement with AI-generated content.

5. Detection Capability

Can you detect whether content in your pipeline was AI-generated (either your own or third-party content)?

No Detection (Gap)

We have no tools or processes to detect AI-generated content entering or leaving our pipeline.

Manual Review

Human reviewers inspect content but no automated detection tools are deployed.

Automated Detection (Compliant)

Automated tools verify watermarks and metadata. Detection is integrated into the content publishing workflow.

6. Organizational Process

Do you have a documented internal process for AI content marking compliance, including roles, responsibilities, and audit procedures?

No Process (Gap)

No documented process exists. AI content marking is ad-hoc or non-existent.

Informal Guidelines

Some internal guidelines exist but they are not formally documented, assigned to owners, or audited.

Documented Process (Compliant)

Formal policy with assigned roles, documented procedures, and periodic audit/review schedule.

Article 50 Content Marking Assessment Record

Disclaimer: This assessment uses the Article 50 legal baseline plus current second-draft code best practices as of April 20, 2026. The final Code may differ. This tool does not constitute legal advice. Consult qualified counsel for binding compliance decisions.

AI Content Marking Compliance Checker