Starting August 2, 2026, your organization must disclose AI system use to individuals under Article 50 of the EU AI Act. This template provides the necessary disclosure language, technical implementation notes, and validation criteria to meet transparency obligations before the deadline.
Purpose of the Template
Article 50 outlines four transparency obligations based on whether you're a provider or deployer and the function of your AI system. This template focuses on the most common requirement: informing users they're interacting with an AI system (Article 50(1)) and disclosing AI-generated or AI-manipulated content (Article 50(4)).
Use this template when:
- Your chatbot, voice assistant, or automated agent interacts directly with customers or employees.
- You publish AI-generated text that informs the public on matters of public interest.
- You deploy emotion-recognition or biometric-categorization systems.
- You generate or manipulate images, audio, or video depicting real persons, objects, places, or events.
The template provides baseline disclosure language, which you'll customize based on your system's function, audience characteristics, and where the disclosure appears in your user journey.
Prerequisites
Before using this template, ensure:
Legal scoping complete: Identify which AI systems trigger Article 50 obligations. If relying on an exemption (such as the "obviousness" exception), document your analysis.
Roles assigned: Determine if you're the provider, deployer, or both. If using third-party AI tools, ensure vendor contracts specify who handles disclosure implementation.
Integration points identified: Decide where disclosures will appear, onboarding screens, content headers, user interfaces, published materials, or physical signage.
Accessibility requirements understood: Assess if your audience includes vulnerable individuals or children who need adapted disclosure language.
GDPR alignment checked: Review how Article 50 disclosures interact with your existing General Data Protection Regulation transparency obligations, especially if your AI system processes personal data.
The Template
User Interaction Disclosure (Article 50(1))
Context: Display this when users first interact with your AI system, unless AI use is obvious.
Standard disclosure: "You are interacting with an AI system. This [chatbot/assistant/tool] uses artificial intelligence to [describe primary function]. [Optional: Human oversight is available by [method].]"
Simplified disclosure for vulnerable users: "This is a computer assistant. It uses AI to help you. A person can help you instead if you ask."
Technical implementation note: Embed this disclosure in your system's initial response, welcome screen, or first interaction. For voice systems, include it in the opening greeting. For text-based systems, display it prominently before the user submits their first query.
AI-Generated Content Disclosure (Article 50(4))
Context: Use this when publishing AI-generated or AI-manipulated text intended to inform the public on matters of public interest.
Disclosure for published content: "[AI-Generated Content] This [article/report/summary/translation] was created using artificial intelligence. [Optional: It has been reviewed by [role] before publication.]"
Disclosure for manipulated content: "[AI-Manipulated Content] This [document/image/recording] was modified using artificial intelligence. [Describe nature of modification: e.g., 'The original text was translated and summarized' or 'Background elements were removed from the image.']"
Technical implementation note: For web content, include disclosure as a visible header or footer element. For documents, embed it in metadata and display it prominently on the first page. Consider using the European Commission's voluntary transparency icons alongside text disclosures.
Deepfake Disclosure (Article 50(4))
Context: Required when you generate or manipulate image, audio, or video content depicting a person, object, place, or event in a way that could mislead.
Standard disclosure: "[Synthetic Media] This [image/video/audio] depicts [subject] and was artificially generated using AI. It does not represent a real event or unaltered recording."
Artistic/creative work disclosure: "This work contains AI-generated imagery for [artistic/satirical/fictional] purposes."
Technical implementation note: Embed machine-readable markers in the content itself (metadata, watermarks, or cryptographic provenance mechanisms). Display human-readable disclosure adjacent to the content when published.
Emotion Recognition Disclosure (Article 50(3))
Context: Required when deploying emotion-recognition or biometric-categorization systems.
Physical environment disclosure: "Notice: This area uses AI systems that analyze [facial expressions/biometric characteristics]. [Purpose of processing]. Your participation is [required/voluntary]. For more information, see [privacy notice location]."
Digital environment disclosure: "This application uses AI to analyze [emotion/biometric data] for [stated purpose]. You can [opt out/request human review] by [method]. See our privacy notice for details."
Technical implementation note: For physical spaces, display signage at entry points. For digital systems, show disclosure before data collection begins. Ensure your disclosure satisfies both Article 50 and General Data Protection Regulation transparency requirements.
Customization Guidelines
Adjust for audience sophistication: If your users are technical professionals, reference AI model types or capabilities. For consumers or vulnerable populations, use plain language and define technical terms.
Integrate with existing notices: Fold AI disclosures into your privacy notice, terms of service, or product documentation. Users prefer consolidated information.
Calibrate prominence: Ensure disclosures are "clear, accessible and provided at the appropriate point in the user journey." For high-stakes interactions, make disclosures more prominent.
Specify opt-out or human review options: Where feasible, inform users how to request human intervention or decline AI-based processing. This strengthens compliance with both Article 50 and General Data Protection Regulation rights.
Document your "obviousness" analysis: If you conclude disclosure isn't required because AI use is obvious, document your reasoning. Consider whether a reasonable person would recognize they're interacting with AI.
Add technical markers for generated content: Article 50(2) requires machine-readable detection for synthetic media. Implement watermarking, metadata tagging, or provenance tracking in addition to human-readable disclosures.
Validation Steps
Pre-deployment review: Before launching any AI system or publishing AI-generated content, confirm:
- Disclosure language is present and correct.
- Technical markers are embedded.
- Disclosure appears at the right point in the user journey.
- Language is appropriate for your audience, including vulnerable users.
User testing: Show your disclosure to representative users and ask: "Do you understand you're interacting with AI?" and "Do you know what the AI system does?" Adjust if responses suggest confusion.
Cross-reference with vendor contracts: If using third-party AI systems, verify your vendor has implemented their provider-side obligations.
GDPR alignment check: Review your Article 50 disclosures against your General Data Protection Regulation Article 13/14 information requirements. Ensure you're not duplicating or contradicting obligations.
Exemption documentation: If relying on any exemption, maintain written justification. Regulators will expect you to explain why Article 50 doesn't apply.
Periodic review: Set a calendar reminder to review your disclosures quarterly. AI system capabilities change, use cases evolve, and regulatory guidance develops. Your February 2026 disclosure may need updating by August 2026.
Regulators will assess not just whether you disclosed, but whether your disclosure was effective. Test your implementation with real users, document your decisions, and update your disclosures as your AI systems change.
