The Code of Practice was developed through a multi-stakeholder process facilitated by the European AI Office. While participation is voluntary, the European Commission and the AI Board have recognised it as an appropriate framework for demonstrating compliance with the AI Act's transparency obligations. Organisations choosing not to follow the Code will need to demonstrate compliance through other means.
- Access the Code of Practice on Transparency of AI-generated Content
- Access the Guidelines on transparency obligations for providers and deployers of AI systems
The Code focuses on the transparency obligations set out in Article 50 of the AI Act, which aim to reduce the risks of deception, manipulation and misinformation associated with AI-generated content. Among other measures, the Code:
- Introduces optional standardised icons for labelling AI-generated and AI-manipulated content.
- Supports machine-readable marking of AI-generated content;
- Promotes effective, interoperable and reliable methods for identifying AI-generated materials.
- Requires disclosure when deepfake content has been generated or manipulated using AI.
- Encourages internal compliance procedures and human oversight mechanisms.
Who does it apply to?
The transparency obligations apply not only to AI providers developing and placing AI systems on the market, but also to AI deployers, meaning people using AI systems in a professional capacity. This includes organisations that use AI-generated or AI-manipulated content as part of their activities. Personal use by private individuals is not covered by these provisions.
For museums and other cultural organisations, this means that transparency obligations may apply when using AI tools to create or modify content for exhibitions, websites, social media, publications or other public-facing activities. However, the guidance also clarifies that certain AI-generated text relating to matters of public interest may be exempt from disclosure requirements if it has undergone appropriate human review or editorial control and a responsible editor has assumed responsibility for the publication.
Where AI-generated or AI-manipulated images, audio or video are used, these should be clearly labelled as such. The new optional icons published alongside the Code can help organisations communicate this transparently.
