The Transparency Deadline Is Weeks Away
While much attention has focused on the high-risk system obligations taking effect on 2 August 2026, the same deadline brings another critical set of requirements into force: the Article 50 transparency rules. These provisions apply to a much broader set of AI systems than the high-risk framework, and many organizations are unprepared.
Article 50 of the EU AI Act establishes transparency obligations for providers and deployers of certain AI systems. Unlike the high-risk regime, which applies to specific use cases listed in Annex III, the transparency rules apply to any AI system that interacts with humans, generates content, or performs biometric analysis. This means companies using chatbots, generating marketing content with AI, or deploying emotion recognition tools all need to comply.
The European Commission is preparing dedicated guidelines on the practical application of Article 50, expected as part of the broader 2026 guidance package. But with the enforcement deadline approaching, organizations cannot afford to wait for final guidance before preparing.
What Article 50 Requires
Article 50 creates three distinct transparency obligations, each targeting a different type of AI interaction.
1. Disclosure of AI Interaction
Providers must ensure that AI systems designed to interact with natural persons are designed and developed in such a way that individuals are informed that they are interacting with an AI system. This applies to chatbots, virtual assistants, conversational agents, and any other system where a person might reasonably believe they are communicating with a human.
The obligation falls on providers to design the disclosure mechanism into the system. Deployers must ensure that the disclosure remains active during use and is not disabled or circumvented.
2. Marking of AI-Generated Content
Providers of AI systems that generate synthetic audio, image, video, or text content must ensure that outputs are marked in a machine-readable format as artificially generated or manipulated. This requirement exists alongside the EU’s broader approach to combating disinformation and ensuring that individuals can distinguish between authentic and AI-created media.
The key technical question is what constitutes a “machine-readable” mark. The Commission is working on a Code of Practice for the marking and labelling of AI-generated content, which will provide implementation guidance. Until this code is finalized, providers should implement available content provenance standards, such as C2PA (Content Authenticity Initiative) or similar technical watermarking.
3. Disclosure of Deep Fakes and AI-Generated Public Interest Content
Deployers of AI systems that generate or manipulate image, audio, or video content that constitutes a deep fake must disclose that the content has been artificially created or manipulated. This obligation applies broadly to any deployer who publishes or shares such content.
The disclosure requirement is stricter for content concerning matters of public interest. In such cases, deployers must ensure that the disclosure is clear and prominent, allowing the audience to understand that they are viewing AI-generated content.
Exceptions apply where the content is part of evidently artistic, creative, fictional, or satirical work, provided that this is evident from the context. Law enforcement use is also exempted under specific conditions.
Who Must Comply
The transparency rules apply to a wider set of organizations than many companies realize. The following table clarifies who bears obligations:
| AI System Type | Provider Obligation | Deployer Obligation |
|---|---|---|
| Chatbots and conversational AI | Design disclosure into system | Ensure disclosure is active during use |
| Generative AI for content creation | Mark outputs as machine-readable AI-generated | Label deep fakes and public interest content |
| Emotion recognition systems | Inform individuals they are subject to emotion recognition | Ensure notification occurs during deployment |
| Text generation for public interest | Mark outputs as AI-generated | Disclose AI authorship when publishing |
Importantly, these obligations apply regardless of the risk classification of the underlying AI system. A chatbot built on a general-purpose AI model may not be high-risk under Annex III, but it still triggers Article 50 transparency requirements.
Common Compliance Gaps
Based on the AI Office stakeholder consultations and preliminary guidance, several compliance gaps have emerged.
Gap 1: Customer Service Chatbots Without Disclosure
Many organizations deploy chatbots on their websites or in customer service channels without clear disclosure that users are interacting with AI. Even when terms of service mention AI use, this does not satisfy the requirement for real-time, contextual disclosure during the interaction.
Compliance requires visible, prominent disclosure at the point of interaction. A general statement buried in a privacy policy will not be sufficient.
Gap 2: AI-Generated Marketing Content Without Marking
Marketing teams increasingly use generative AI to create blog posts, product descriptions, social media content, and advertising copy. Under Article 50, outputs from these systems must be marked in a machine-readable format. Many organizations are not yet implementing content provenance standards for marketing materials.
Gap 3: Internal AI Tools Without Employee Disclosure
Article 50 applies to interactions with natural persons, including employees. If an organization deploys internal AI tools that interact with employees, such as HR assistants or IT support bots, employees must be informed they are interacting with AI.
Gap 4: Third-Party AI Systems
Many organizations deploy AI systems built by third parties. Deployers must understand whether their providers have implemented the required disclosure mechanisms. If the provider has not, the deployer bears the compliance gap and potential liability.
Practical Implementation Steps
For AI System Providers
-
Implement interaction disclosure: Build clear, visible disclosure mechanisms into any AI system that interacts with users. The disclosure should be unambiguous and appear at the start of or during the interaction.
-
Adopt content provenance standards: Implement machine-readable watermarking or metadata tagging for all AI-generated outputs. C2PA, SynthID, or equivalent standards provide technical frameworks for compliance.
-
Document implementation choices: Record how your system satisfies transparency requirements, including the technical standards used for content marking. This documentation will be essential if authorities request evidence of compliance.
-
Provide deployer instructions: Clearly inform deployers of their Article 50 obligations, including how to maintain disclosure settings and label generated content.
For AI System Deployers
-
Audit your AI touchpoints: Identify every AI system in your organization that interacts with natural persons, generates content, or performs emotion recognition. Include customer-facing systems, internal tools, and marketing workflows.
-
Verify provider compliance: For each third-party AI system, confirm that the provider has implemented the required transparency features. Request documentation of their compliance approach.
-
Establish content labeling workflows: Create processes for labeling AI-generated content before publication. Define who is responsible for applying labels and how compliance is verified.
-
Train customer-facing teams: Ensure that employees managing AI-powered channels understand how disclosure works and what to do if the disclosure mechanism fails.
What the Code of Practice Will Add
The AI Office has selected a Code of Practice for the marking and labelling of AI-generated content. Once finalized, this voluntary code will provide specific technical guidance on:
- How to implement machine-readable content marking
- Which technical standards satisfy the marking requirement
- How to design user interfaces for disclosure
- How to handle edge cases, such as AI-assisted content that is edited by humans
While the code is voluntary, organizations that adopt it benefit from a presumption of conformity with Article 50 requirements. Following the code reduces regulatory risk and simplifies compliance audits.
Penalties for Non-Compliance
Failure to comply with Article 50 falls under the general penalty provisions of Article 99. Organizations that violate transparency obligations face fines of up to €7.5 million or 1% of global annual turnover, whichever is higher. While these penalties are lower than those for high-risk system violations, they are still substantial enough to warrant serious compliance attention.
More importantly, the reputational risk of non-compliance with transparency rules can be severe. Consumers and business partners increasingly expect honesty about AI use. Organizations that fail to disclose AI interactions risk losing trust and credibility.
The Broader Transparency Landscape
Article 50 does not exist in isolation. Organizations must consider how AI transparency obligations interact with other regulatory frameworks:
-
GDPR: Data subjects have rights to information about automated decision-making (Article 22). AI transparency under the AI Act complements but does not replace these rights.
-
Digital Services Act: Online platforms face additional content labeling requirements under the DSA. AI-generated content may trigger both regimes.
-
AI Office guidelines on interplay: The Commission will publish joint guidance with the European Data Protection Board on how the AI Act and GDPR interact, including on transparency.
Organizations should take a holistic approach to transparency, ensuring that their disclosures satisfy all applicable regulatory frameworks rather than treating each as a separate compliance exercise.
Conclusion
The Article 50 transparency deadline arrives on 2 August 2026, alongside the broader high-risk system enforcement. For many organizations, transparency compliance is more immediately relevant than high-risk obligations, because it applies to common AI use cases like chatbots and content generation. Providers and deployers that start implementing disclosure mechanisms now will avoid a last-minute scramble as the deadline approaches.
The key actions are straightforward: audit your AI touchpoints, implement technical marking standards, establish labeling workflows, and train your teams. The legal requirements are clear, even as specific implementation guidance continues to evolve.
Jurista.ai helps organizations map their Article 50 transparency obligations, identify compliance gaps, and implement the technical and organizational measures needed before the August 2026 deadline. Assess your compliance at jurista.ai.