FIELD GUIDEEU AI ACT
Make the rule easier to understand.
Plain-language notes for the decisions behind an AI transparency review. Start with scope, follow the evidence, and keep human judgment in the loop.
ARTICLE 50 · OVERVIEW
Article 50: roles, disclosures and scope
Understand who has each duty, which disclosure applies, and why a single scan cannot decide legal scope.
Read guideARTICLE 50(1) · INTERACTION
AI interaction notices for chatbots
Review when and where people need to be told they are interacting with an AI system.
Read guideARTICLE 50(2) · PROVENANCE
C2PA and machine-readable provenance
Learn what provenance markers can show—and what they cannot prove about authenticity.
Read guideARTICLE 50(2) + (4) · MEDIA
Synthetic media: provider and deployer duties
Separate provider marking from deployer disclosure, with context and exceptions in view.
Read guideENFORCEMENT · CONTEXT
Transparency enforcement and penalties
Understand the enforcement context without treating a technical observation as a legal risk estimate.
Read guideSYSTEM ROLES · SCOPE
Open-weight models in public applications
Distinguish an open-weight model from the public-facing AI system built with it.
Read guideARTICLE 50(4) · DISCLOSURE
Deepfake and voice-clone disclosures
Review deepfake disclosure scope, artistic context and the limits of automated media checks.
Read guideARTICLE 50(1) · IMPLEMENTATION
Implementing a clear AI notice
Start with a visible interaction notice, then test its timing, language and mobile presentation.
Read guidePRIVACY · AI DISCLOSURES
AI notices and privacy policies
Keep privacy explanations separate from notices people need at the point of interaction.
Read guideVENDOR REVIEW · CHECKLIST
Questions to ask an AI vendor
Ask vendors for role clarity, evidence, disclosure behavior and change-handling details.
Read guide