EU Publishes Voluntary Code for General-Purpose AI to Guide Responsible Innovation

In July, the European Commission unveiled the final version of the Code of Conduct for General-Purpose AI (GPAI), aiming to help companies align with forthcoming obligations under the AI Act. Although adoption is voluntary, the code is designed to become a market benchmark for developers of foundational models. 

The GPAI Code is built around three core pillars: transparency, copyright compliance, and safety. Under transparency, model creators must maintain detailed, standardized documentation to ensure clarity for both users and supervisory bodies. Copyright compliance emphasizes responsible data use and risk management in model training. The safety aspect primarily targets advanced AI providers who could present systemic risks and lays out special obligations under Article 55 of the AI Act. 

Proponents argue that the code provides practical guidance and helps reduce legal uncertainty. Critics caution that compliance costs, especially for smaller firms, risk stifling innovation. The challenge ahead: striking a balance between protecting users and enabling AI-driven growth.

(PWC)