The European Commission has published its final voluntary code for marking and labeling AI-generated content, giving brands and platforms a practical framework ahead of the AI Act's transparency obligations. The Commission's announcement describes measures for machine-readable marking by model providers and disclosure by deployers when synthetic or manipulated material could be mistaken for authentic content.
The code itself is not the law. It is a compliance aid tied to Article 50 of the AI Act, and the Commission's policy page explains how it supports the transparency requirements. A Jones Day analysis notes that the Commission released the final code in June 2026 as organizations prepared for the obligations to apply.
The hard part is provenance, not the badge
A visible label is the last step in a longer chain. Retailers increasingly use generative tools for background extension, product visualization, localization, copy variation, video, and personalized creative. If information about those edits is lost between the model, agency, digital-asset manager, ad server, marketplace, and platform, the brand may be unable to apply the right disclosure consistently.
That makes this a workflow-design problem. Teams need to decide what counts as generated or materially manipulated, attach machine-readable metadata where required, preserve it through exports, and define when a consumer-facing label is triggered. Contracts with agencies and production vendors should specify who records the provenance data and who is responsible for final placement.
Avoid both overlabeling and omission
Label every routine assist indiscriminately and disclosures can become noise. Label too little and a realistic synthetic scene, person, voice, or event can create legal and trust risk. The Commission's framework is intended to create more consistent practices, but brands still need documented judgment for their own use cases. Product imagery deserves special care because the line between harmless cleanup and a materially altered representation can affect both AI transparency and consumer-protection rules.
Global campaigns should not assume that platform labels alone close the gap. Meta, Google, TikTok, publishers, and retailers can apply different policies, while the advertiser retains responsibility for its claims and creative process. The safest operating model stores one authoritative provenance record with the asset and then maps channel-specific disclosures from it.
The deadline should not be treated as a last-minute creative review. Audit active tools and vendors now, classify common use cases, test whether metadata survives the production stack, and create a human escalation path for realistic synthetic content. Europe's rule turns AI disclosure from a copy edit into infrastructure.