EU AI Act Article 50 Is Live: What Providers and Deployers Need to Make Transparent in 2026
Published August 2026 by Solarc Labs
Article 50 transparency obligations apply from 2 August 2026
The European Commission's July 2026 guidelines state that the AI Act transparency obligations in Article 50 apply from 2 August 2026. They are not one generic “AI label” rule. Different duties attach to different actors and system or content types, so the first implementation step is to identify whether the organisation is acting as a provider, a deployer, or both for the surface in question. That role map matters because provider-side system design duties and deployer-side disclosure duties are not interchangeable. A single disclosure badge cannot safely stand in for the complete Article 50 analysis.
Providers have interaction and machine-readable marking duties
Article 50 requires providers of AI systems intended to interact directly with natural persons to ensure people are informed that they are interacting with an AI system unless that is obvious in the circumstances. Providers of systems that generate synthetic audio, image, video or text also have a machine-readable marking obligation for generated or manipulated outputs, subject to the legal scope and exceptions. For product teams, this creates both UX and technical work: the interaction disclosure has to be clear at the right moment, while output marking needs a reliable implementation that survives the actual content pipeline.
Deployers have separate disclosure duties for specific uses
The deployer side includes disclosure obligations around emotion-recognition and biometric-categorisation systems, deepfakes and certain AI-generated or manipulated text published to inform the public on matters of public interest. The exact scope and exceptions matter, including the treatment of human review and editorial responsibility for some public-interest text. That means a publisher, agency or application operator should not start with “do we use generative AI?” and stop there. It should identify the content or interaction surface, the organisation's role and which Article 50 paragraph actually applies.
Machine-readable provenance is evidence, not the whole legal test
The Commission's transparency code addresses marking and detection of AI-generated content and encourages interoperable, robust machine-readable approaches. Media provenance technologies such as C2PA can be relevant evidence in a content pipeline, but Article 50 does not simply say “add C2PA and you are compliant.” The legal obligation, technical marking method, visible disclosure and applicable exceptions still need to be evaluated together. Operationally, preserve what the file actually says: credentials present, credentials absent, verification failed, or provenance unknown. Do not convert an unknown provenance state into a claim that content is human-made.
Build a disclosure inventory before redesigning every surface
List the AI-enabled user interactions and content-generation paths in scope. For each one, record the provider/deployer role, affected content or system type, current user-facing disclosure, machine-readable marking mechanism if relevant, accessibility treatment, editorial review state and owner for legal interpretation. That inventory reveals where the organisation has a real implementation gap versus where a disclosure already exists but lacks evidence or consistency. The result should be a review queue tied to real surfaces, not a generic policy document that cannot be traced back to the product.
AI Transparency Pack is provenance review, not compliance certification
AI Transparency Pack can help inspect supported media for provenance evidence and keep unknown or unverifiable states explicit. It is not an AI detector and does not certify compliance with Article 50. It also cannot decide on its own whether a particular provider or deployer falls within an obligation or exception. Use the Article 50 guidelines and the current Commission code as the legal-policy reference, then use the provenance-review guide when the narrower job is to inspect and preserve machine-readable media evidence without overstating what that evidence proves.
Primary sources
Sources used for this article
Continue the job