On 2 August 2026, the transparency obligations set out in Article 50(1), (2) and (5) of the EU AI Act become applicable. Following the publication of the European Commission's guidelines and supporting FAQs, providers placing AI systems on the EU market must ensure compliance with these requirements before placing a system on the market or putting it into service.
Article 50 also contains obligations for deployers of certain AI systems, including requirements relating to deep-fake disclosures, AI-generated public-interest content, emotion recognition systems and biometric categorization systems. Please note this is non-binding guidance and that only the CJEU can provide authoritative interpretation.
Providers of AI systems that interact directly with natural persons, including chatbots, AI assistants, AI agents and similar systems, must ensure that users are informed that they are interacting with an AI system, unless this is obvious from the context. The Commission has clarified that this exception should be interpreted restrictively.
Providers of generative AI systems that fall within the scope of Article 50(2) and generate or manipulate synthetic text, images, audio or video must also ensure that AI-generated or AI-manipulated outputs are marked in a machine-readable format and remain detectable as artificially generated or manipulated. Appropriate technical measures should be incorporated during the design and development phase and supported by relevant technical documentation and evidence of compliance.
The Commission's guidance reinforces that transparency requirements must be addressed as part of the AI system lifecycle and should not be treated as post-market disclosures. Providers should therefore review product portfolios, technical controls and governance processes to ensure Article 50 requirements are met ahead of the application date.
The objective is to increase transparency and trust in both AI-generated content and AI-human interactions.
When Do the Requirements Apply?
|
Requirement
|
Applicability Date
|
|
Article 50 transparency obligations
|
2 August 2026
|
|
Grace period for certain AI-generated content marking obligations for systems already on the market before 2 August 2026
|
Until 2 December 2026
|
Providers placing new AI systems on the EU market after 2 August 2026 need to comply with these requirements.
Key Requirements for providers
AI-Human interaction disclosure
|
Requirement under Article 50
|
What providers need to ensure
|
|
Disclosure of AI interaction
|
Providers of AI systems that interact directly with natural persons, such as chatbots, AI assistants, AI agents and virtual avatars, must ensure that users are informed they are interacting with an AI system, unless this is obvious from the circumstances. The Commission notes that the "obviousness" exemption should be interpreted restrictively.
|
|
Timing of disclosure
|
Information that a user is interacting with AI should be provided from the beginning of the first interaction, enabling individuals to make informed decisions regarding their engagement with the system.
|
|
Manner of disclosure
|
The notification should be clear, distinguishable, understandable and accessible to the intended users. The objective is to ensure transparency and avoid any ambiguity regarding whether the interaction is with a human or an AI system.
|
|
Provider considerations
|
Providers should review AI systems that directly interact with users and assess whether existing user interfaces and disclosures satisfy Article 50 requirements. Where reliance is placed on the "obviousness" exemption, providers should be able to justify and document the basis for that determination.
|
AI-Generated Content Marking and Detectability
|
Requirement under Article 50
|
What providers need to ensure
|
|
Marking of AI-generated content
|
Providers of AI systems that generate or manipulate images, audio, video or text must ensure that outputs are marked in a machine-readable format so that they can be detected as artificially generated or manipulated. The marking should be effective, reliable, robust and interoperable.
|
|
Scope of application
|
The obligation applies broadly to generative AI systems, including systems built on general-purpose AI technologies where they generate or manipulate synthetic content.
|
|
Technical implementation
|
Providers should implement appropriate technical measures capable of supporting the marking and detection of AI-generated or AI-manipulated content. The Commission recognizes that implementation should take account of the state of the art, technical feasibility and the characteristics of different content types.
|
|
Evidence of compliance
|
Providers should be able to demonstrate that suitable marking and detection mechanisms have been implemented and maintained. This may include technical documentation, validation activities and records supporting the effectiveness of the chosen solution.
|
Product Design Considerations
Providers should note that the Commission does not view Article 50 transparency obligations as post-market disclosures. Compliance is expected to be built into the AI system prior to placement on the market or entry into service. Providers should therefore ensure that transparency requirements are reflected in product design specifications, development and validation activities, and supporting technical documentation. As part of their compliance preparations, providers may wish to assess whether existing development processes, software supply chains and governance arrangements provide sufficient assurance that Article 50 requirements are consistently implemented and maintained.
Documentation and Evidence
Providers should not only implement the transparency measures required by Article 50 but also maintain sufficient evidence to substantiate that those measures are effective and operating as intended. Where providers choose not to adhere to the recognized Code of Practice, they should be prepared to explain and justify the alternative measures adopted to achieve compliance. Consequently, organizations may wish to review whether their technical documentation, validation activities, record-keeping practices and governance arrangements provide sufficient evidence to support future regulatory scrutiny or market surveillance activities.
Recommendations for providers
With the application of Article 50 approaching on 2 August 2026, providers should review whether any AI systems within their portfolio are subject to the transparency requirements established by the AI Act. Attention should be given to systems that interact directly with natural persons, as well as systems capable of generating or manipulating text, images, audio or video content. The Commission's guidance provides further clarification on the circumstances in which transparency obligations apply and the limited exemptions that may be available.
Providers should also consider whether existing transparency measures adequately support compliance with the requirements relating to disclosure of AI interactions and the marking and detectability of AI-generated content. As these obligations apply before a system is placed on the market or put into service, organizations may need to evaluate whether current design, development and documentation processes provide sufficient evidence of conformity with Article 50.
The publication of the Commission's guidance should also be viewed within the broader context of AI Act implementation. Transparency obligations are closely linked to wider governance, risk management and compliance activities, and further guidance from the European AI Office and national competent authorities is expected as implementation progresses.
Article 50 represents a significant shift from voluntary transparency practices to enforceable regulatory obligations. Providers should treat transparency controls as product compliance requirements rather than communications features.
We recommend to:
- Establish clear accountability for Article 50 compliance.
- Incorporate transparency requirements into AI governance frameworks.
- Verify technical controls for AI-generated content marking and detection.
- Prepare objective evidence to support future regulatory inspections and market surveillance activities.
Organizations that act now will be better positioned to demonstrate compliance, reduce enforcement risk, and build trust with customers and regulators as the AI Act enters its next implementation phase.
For additional information you can contact us at AIBSIContactPage@bsigroup.com.
If you're already a BSI client, contact your Scheme Manager.
Sincerely,
Linda Moon
Director of Global Quality & Accreditation
Regulatory Services, BSI
Yoann Buisson
Associate Head of Global Quality and Accreditation
Regulatory Services, BSI
