Research & Alerts This is a sample of a Nymity Research Regulatory Insight & Alert

The guidelines confirm two categories of obligations that apply from August 2, 2026:

• providers of interactive AI systems (e.g., chatbots and virtual assistants) must clearly inform individuals that they are interacting with an AI system, unless this is obvious from the circumstances and context of use; and
• providers of generative or manipulative AI systems must implement machine-readable markings and technical detection mechanisms capable of identifying AI-generated or manipulated audio, image, video, or text output as artificially created:
  • markings must be technically robust, interoperable, effective, and reliable to the extent technically feasible; and
  • detection mechanisms must account for the type of content and the sector-specific and technical limitations involved.

How the disclosure obligations differ by AI system category

• deployers publishing or distributing deep fake content must disclose that the content has been artificially generated or manipulated, unless it forms part of an evidently artistic, creative, satirical, fictional, or analogous work, subject to appropriate safeguards for third-party rights;
• deployers of AI systems generating or manipulating text published to inform the public on matters of public interest must disclose that the text has been artificially generated or manipulated, unless the content has undergone human review or editorial control and a natural or legal person holds editorial responsibility for its publication; and
• deployers of emotion recognition systems or biometric categorisation systems must inform individuals exposed to the system of its operation, and, where personal data is processed, must do so in accordance with the GDPR.

Additional guidance clarified in the final text

• disclosures must be provided clearly and distinguishably at the latest at the time of first interaction or exposure;
• disclosures must be accessible to persons with disabilities, in accordance with applicable EU accessibility requirements; and
• special consideration must be given to vulnerable individuals, including by reason of age or disability, and to children.

The final guidelines are largely consistent with the draft version published for consultation, with the EC clarifying the interplay between Article 50 and the AI Act’s prohibited-practice and AI-literacy provisions (Articles 5 and 4) and adding worked examples for interactive AI and synthetic media use cases.

Primary transparency obligations clarified under the final guidelines.

The EC’s final text incorporates feedback received during the public consultation on the draft guidelines, most notably by narrowing the human-review exemption for public-interest text disclosure and adding worked examples distinguishing in-scope chatbots from simple rule-based customer service tools.

The guidance also confirms these are the EC’s own non-binding guidelines rather than a formally adopted implementing or delegated act, meaning national courts and the EU AI Office retain interpretive authority, though the guidelines are expected to carry significant persuasive weight for national competent authorities and market surveillance authorities enforcing Article 50 from August 2, 2026.

Distributors, importers, and integrators that place or make available AI systems in scope of Article 50 are advised in the guidelines to confirm, contractually where necessary, which party in the supply chain bears each disclosure obligation before the application date.

The guidelines also flag that the EC intends to monitor implementation during the first year of application and may issue supplementary technical guidance on machine-readable marking standards once the relevant harmonised standards under the AI Act are finalized.

The guidelines apply to any provider or deployer placing an in-scope AI system on the EU market or putting it into service or use within the EU, regardless of where the provider or deployer is established, where the system’s output is used within the EU.

National competent authorities and market surveillance authorities are expected to reference the final guidelines when assessing compliance with Article 50 from the August 2, 2026 application date, alongside their existing powers under the AI Act’s market surveillance framework.

• Chatbot and virtual-assistant providers — squarely in scope; the guidelines’ worked examples focus heavily on distinguishing genuinely interactive AI from static, rule-based tools.
• Generative AI and synthetic media providers — in scope for the machine-readable marking obligation, including providers embedding third-party generative models into their own products.
• Publishers and platforms distributing deep fakes or AI-assisted public-interest content — in scope as deployers, independent of whether they also develop the underlying model.
• Organizations using emotion recognition or biometric categorisation in the workplace, marketplaces, or public spaces — in scope for the deployer notice obligation, and should also confirm the use does not fall within an Article 5 prohibited practice.

Related source documents referenced in the guidelines include the EC’s final guidance text, the earlier draft guidelines published for consultation, the AI Act’s official consolidated text (Articles 4, 5, 50, and 55), and the European Commission’s accompanying summary of stakeholder feedback received during the consultation period.

Nymity Research tracks each of these documents as they are published, along with the EC’s broader AI Act implementation guidance and the EU AI Office’s enforcement priorities, so subscribers can see how this guidance fits the Act’s wider implementation timeline rather than reading it as an isolated release.