Pulsantiera di navigazione Home Page
Pagina Facebook Pagina Linkedin Canale Youtube Italian version
News
Legal news

INFORMATION TECHNOLOGY

EU Commission: the second draft of the European Code of Conduct on the marking and label-ling of generative AI-generated content has been published.

To help providers and deployers meet the marking and labeling requirements for AI-generated content under the Article 50 AI Act, the Commission is facilitating the development of a voluntary code of conduct.

The most recent draft of the code has been simplified, providing more flexibility, reducing compliance burdens, and incorporating additional technical considerations to improve legal clarity and practicality. It promotes the use of open standards for AI-powered content reporting and an EU icon for labeling, in order to simplify compliance and reduce costs.

The draft code of conduct for the transparency of AI-generated content is structured in 2 sections, each dealing with different aspects of transparency and regulation for providers and deployers, respectively.

Section 1 covers the marking and detection of AI content and is aimed at providers of generative AI systems under Article 50(2) of EU Regulation 1689/2024 (AI Act). Compared to the first draft, this section of the code has undergone significant changes and has introduced more flexibility and clarity. The current revision removes and consolidates several measures and introduces optional elements, while ensuring that all measures remain technically feasible and proportionate. Key commitments include a revised two-tier marking approach involving secure metadata and watermarks, optional fingerprinting and logging, and protocols for detection and verification.

Section 2, aimed at deployer AI systems, focuses on the labeling of deepfakes and textual publications concerning issues of public interest under Article 50(4) of EU Regulation 1689/2024. Compared to the first draft, this section takes a more flexible and practice-oriented approach. It has been restructured to simplify and simplify engagements, while the taxonomy that distinguished AI-generated content from AI-assisted content has been completely removed. Section 2 now presents design and placement requirements that can be applied to icons, labels, or disclaimers, ensuring a minimum level of consistency and allowing you to devise solutions tailored to your needs. In addition, the section proposes a task force to develop a future, uniform and interactive EU icon.

The Code also defined specific regimes applicable to artistic, creative, satirical and fictional works and textual publications under human review or editorial control, facilitating trust in existing practices or procedures.

The annex to the second draft now includes illustrative examples of a potential EU icon to be made freely available.

Stampa la pagina