European Commission Introduces First Code of Practice for Artificial Intelligence
On July 10, 2025, the European Commission announced the implementation of the first Code of Practice for general-purpose artificial intelligence models. Despite its voluntary nature, it will become a key tool for meeting the requirements of the forthcoming AI Act regulation.
The Code is an important document for companies that develop, distribute, or integrate artificial intelligence models in the European Union. Its primary aim is to ensure transparency, uphold intellectual property rights, and manage risks in the context of artificial intelligence. According to Sergiy Barbashyn, an expert from the working group developing the Code, the document is intended to help EU businesses work effectively with AI without expending significant resources on creating their own policies from scratch.
The Code includes documentation templates that facilitate compliance with requirements, particularly in terms of the lawful use of data and security management. Companies are required to establish risk management systems designed to analyze threats and control their impact. Special attention is given to systemic risk models that could affect the safety and rights of citizens.
The document provides for a one-year grace period of “good faith efforts,” during which incomplete compliance will not be penalized, and includes a transition period until 2027 to allow existing models to adapt to the new requirements.
The latest innovations in AI regulation reflect the EU’s effort to create safer conditions for the development of AI technologies, while still leaving room for their further commercial use. Experts believe this approach could enhance trust in Europe as a leader in the technology sector by blending innovation with effective regulation.
| Key Areas of the Code | Description |
|---|---|
| Transparency | Maintaining documentation on models, architecture, and data with the publication of accessible information |
| Copyright | Using only legally accessible data, considering the signals from rights holders |
| Security and Risk Management | Establishing risk management and reporting systems for systemic risk models |
