The European Union receives the Code of Good Practice for General-Purpose AI

 

This code, developed by a team of thirteen independent experts and enriched with input from more than a thousand stakeholders—including model providers, SMEs, academics, and civil society organizations—marks a crucial milestone ahead of the AI ​​Law's entry into force.

 

The Code is designed to guide industry in complying with the regulations of the AI ​​Act, specifically with regard to the General-purpose AI, whose provisions will be enforceable from the August 2th, 2025The European Commission's AI Office will be responsible for enforcing these rules, first for new models (one year later) and then for existing models (two years later). The main objective is to ensure that all AI models circulating in the European market, even the most powerful ones, comply with strict safety and transparency standards.

 

The document is structured into three fundamental chapters: Transparency and Copyright, applicable to all providers of general-purpose AI models, and Protection and security, relevant to a limited number of suppliers developing the most advanced models.

 

Transparency, Copyright and the Importance of Security

 

Since general-purpose AI models are the basis of numerous AI systems in the EU, the AI ​​Act emphasizes the need for adequate transparency. The AI ​​chapter Transparency The Code provides a standardized, easy-to-use documentation form, making it easier for suppliers to gather the required information in a single location. This will allow for more seamless integration of these models into various products and services.

 

As for CopyrightThe Code offers practical solutions for providers to establish policies that fully comply with current EU legislation, ensuring respect for intellectual property in the development and use of AI.

 

A critical aspect addressed is that of Protection and securitySome general-purpose AI models may carry systemic risks, including potential threats to fundamental rights, security, and even the possibility of facilitating the development of chemical or biological weapons, or the risk of losing control over the model. The AI ​​Law requires providers to assess and mitigate these risks. The Code, in its corresponding chapter, details the more advanced practices for the management of this type of systemic risks.

 

Next Steps and Benefits for Adherents

 

Once the Code is formally endorsed by Member States and the Commission, providers of general-purpose AI models that voluntarily sign it will be able to demonstrate compliance with their obligations under the AI ​​Act. By adhering to the Code, signatories will benefit from a lower administrative burden and a greater legal certainty, compared to those who choose other ways to demonstrate their compliance.

 

To complement this Code, the Commission will publish Additional guidelines on general-purpose AI before the obligations come into force. These guidelines will provide greater clarity on who is included or excluded from the scope of the AI ​​Act's regulations.