The European Union’s Artificial Intelligence Act moves into a new enforcement phase on August 2, when transparency obligations under Article 50 formally take effect across all 27 member states. The European Commission has published official guidelines to support market readiness and ensure uniform enforcement, developed after a public consultation involving member states, the EU AI Board and industry stakeholders.
The guidelines define the practical scope of transparency duties across the digital value chain. Regulators say Article 50 targets risks stemming from generative and interactive AI, including synthetic media, deepfakes, automated impersonation and undisclosed biometric profiling. The obligations aim to protect the digital ecosystem from deception and misinformation by giving citizens the information required to calibrate their trust in online media and automated services.
Providers must design AI systems that interact directly with individuals, such as customer service chatbots, virtual assistants or automated call handlers, so users are explicitly notified upon initial contact that they are communicating with an AI. Providers generating synthetic audio, image, video or text must integrate technical, machine-readable markings and provenance signals, such as digital watermarks or cryptographic metadata, allowing detection tools to identify AI-generated or manipulated outputs.
Deployers using emotion recognition or biometric categorisation must explicitly inform exposed individuals before data processing occurs. Anyone deploying AI to generate or manipulate synthetic media depicting realistic people, places or historical events, so-called deepfakes, must display clear, human-perceivable labels. Media outlets, organisations or digital publishers distributing AI-generated text concerning matters of public interest, such as election updates, public health policies or economic reports, must clearly disclose its artificial origin.
The AI Act, adopted in 2024, is being phased in. Prohibitions on the riskiest practices took effect in February 2025, and rules for general-purpose AI models became applicable in August 2025. From August 2, the EU AI Office and national authorities are responsible for implementing, supervising and enforcing the act. The AI Office holds enforcement powers over general-purpose models: it can request technical documentation, evaluate models, require corrective measures and issue fines for non-compliance. Penalties can reach up to 7 percent of a company’s global annual turnover for the most serious violations.
To satisfy compliance rules, providers and deployers can leverage the voluntary Code of Practice on Transparency of AI-generated Content. Organisations that align with the Code gain a recognised benchmark for legal compliance. Those opting out must independently prove that their proprietary marking and labelling solutions meet equivalent standards of effectiveness, robustness and interoperability. For direct human-AI interaction notices, providers and deployers retain flexibility to implement tailored disclosures adapted to their specific user interfaces.
As enforcement commences, businesses worldwide that serve European consumers are reviewing content workflows, updating conversational UI designs and embedding metadata standards such as C2PA Content Credentials into their software architectures. The milestone marks a major shift for biometrics, deepfake labelling and generative AI compliance, and regulators have signalled that uniform enforcement across the bloc is a priority as the rules begin to bite.
Sources: Identity Week, European Commission, AI Journ
Author: Technology Desk
discussion