EU Releases AI Code of Practice Amid Industry Pushback

The European Commission released a voluntary Code of Practice for general-purpose AI (GPAI) on July 10, 2025, to guide companies through the implementation of the EU AI Act. The framework focuses on three pillars: transparency regarding training data and energy usage, copyright protections—including the exclusion of piracy websites from training sets—and safety and security for models with systemic risks.

While the AI Act's rules for GPAI take effect on August 2, 2025, the EU AI Office will not begin enforcement for new models for at least one year, and existing models have until August 2027 to comply. Companies that voluntarily sign the code may benefit from reduced administrative burdens and increased legal certainty. OpenAI has expressed its intent to sign the agreement once it is finalized.

The initiative has sparked a conflict between industry and civil society. Nearly 50 European firms, including Airbus, ASML, and Mistral AI, signed an open letter requesting a two-year suspension of the Act, arguing that the timeline restricts innovation and global competitiveness. The Commission has rejected these requests for a delay. Conversely, a coalition of 52 civil society organizations, including European Digital Rights and CDT Europe, urged the Commission not to backtrack, warning that simplification could serve as a cover for deregulation and undermine fundamental rights.

Non-compliance with the broader AI Act can result in significant penalties, with some reports citing fines up to 35 million euros or 7% of global revenue, while others note penalties up to 15 million euros or 3% of global revenue.