Europe’s AI labelling rules advanced with the release of a voluntary code on June 10, setting a timeline for mandatory transparency requirements by August 2, 2026. The European Commission’s Code of Practice details how AI-generated content must be marked, focusing on deepfakes, synthetic text, and other materials that might deceive users. While following the code is optional, the obligations it imposes are mandatory.
Providers and Deployers Face Dual Responsibilities
The code divides accountability between developers and users of AI systems. Developers, defined as entities creating AI systems, must implement detection tools that work across platforms by February 2, 2027. These tools must identify AI-generated content and ensure that machine-readable provenance markings remain intact after processing. Users of AI in professional settings must label two categories: deepfakes and AI-generated text on public interest matters. The EU introduced a basic label, a “fully AI-generated” variant, and a “partially AI-modified” version to help with disclosure, though their use remains optional.
Related: Top students celebrate first-class degrees
Some exceptions apply. Artistic, satirical, or fictional deepfakes have lighter disclosure rules. Text reviewed by humans and published under editorial responsibility is not required to be labeled. Law enforcement content also avoids labeling. However, fully automated AI outputs, like news summaries, must use the “fully AI-generated” icon, requiring clear identification.
Timing and Enforcement Challenges
The August 2, 2026, deadline applies to AI systems released after that date. A transitional period runs until December 2, 2026, for systems launched earlier. However, compliance teams must prepare for the original deadline. The Commission and AI Board will evaluate the code’s effectiveness, providing signatories a recognized compliance pathway.
Related: BWY and Parkinson’s UK launch accessible yoga
During drafting, friction arose. The voluntary nature of the code may not fully address these concerns, but the mandatory obligations under the AI Act leave little room for negotiation.
Implications Across Industries
For cybersecurity and eDiscovery providers, the code creates both risks and opportunities. Companies using generative AI in platforms must ensure provenance markings remain intact through data workflows. Litigation teams may use digitally signed metadata to verify disputed evidence, while information governance policies must treat AI transparency data as records, not digital waste.
Related: Using a CV maker to get a well-crafted professional CV
Marketing teams face narrower but real exposure. Synthetic spokespeople, AI-altered imagery, and chatbots must be labeled, but routine AI-assisted ads are generally exempt. Newsrooms receive a carve-out: human-reviewed content avoids labeling, but fully automated outputs require the “fully AI-generated” icon. Third-party validators could use detection tools to verify AI origins, a capability currently missing in fact-checking.
As of June 2026, the code is open for signature, with the August deadline unchanged under current guidance. The challenge for organizations is no longer whether to label, but how to do it effectively—and which labels will be trusted by regulators, clients, and courts. The race to label, detect, and preserve provenance information has begun, with the EU’s transparency rules reshaping AI’s role in business and media.