European Parliament working group welcomes GPAI Code of Practice finalisation
The code will be complemented by guidelines, clarifying which models fall within the scope of the AI Act’s GPAI rules. The AI Act’s obligations on GPAI models will enter into force on 2 August 2025.
The co-chairs of the European Parliament’s working group on the AI Act’s implementation have welcomed the finalisation of the General-Purpose AI (GPAI) Code of Practice, describing it as a timely and practical framework to support responsible AI development ahead of the AI Act’s application date on 2 August 2025.
Brando Benifei (S&D, Italy) and Michael McNamara (Renew, Ireland), who co-chair the Parliament’s Working Group on the Implementation and Enforcement of the AI Act, emphasised the importance of the code as the first step in managing risks associated with powerful AI models. They acknowledged that the drafting process was conducted under considerable pressure, shaped by lobbying and broader geopolitical dynamics, yet still succeeded in striking a balance between industry concerns and civil society demands, particularly around transparency, risk monitoring, and the protection of fundamental rights.
The MEPs called on all providers of general-purpose AI models to sign and implement the code in full, describing participation as a demonstration of good faith and commitment to European values. They warned that voluntary adherence should not be symbolic but should translate into tangible practices. Providers that sign the code will benefit from early regulatory clarity and cooperation with the AI Office.
Highlighting the importance of enforcement, the co-chairs stressed that the code’s effectiveness will depend on the AI Office’s ability to monitor implementation and take appropriate action where necessary. They urged the Commission, member states, and all stakeholders to maintain their commitment to the code and resist efforts to dilute its provisions.
The code of practice, developed by a panel of independent experts and submitted to the Commission on 10 July 2025, is part of a broader set of measures supporting the implementation of the AI Act. While adherence is voluntary, the code serves as a tool for demonstrating compliance with key GPAI obligations. It is complemented by official guidelines clarifying which models fall within the scope of the AI Act’s rules.
