TLDR: In July 2025, the European Commission introduced three crucial instruments—Guidelines on obligations, a voluntary Code of Practice, and a Template for Public Summary of Training Content—to support the responsible development and deployment of General-Purpose AI (GPAI) models. These measures aim to facilitate compliance with the landmark AI Act, which saw its rules for GPAI models become applicable in August 2025, ensuring transparency, safety, and accountability in the rapidly evolving AI landscape.
The European Commission has taken a significant step towards fostering trustworthy artificial intelligence by introducing three key instruments in July 2025, designed to support the responsible development and deployment of General-Purpose AI (GPAI) models. This initiative comes as the rules for GPAI models under the pioneering AI Act became applicable on August 2, 2025, marking a new era for AI governance in Europe.
GPAI models, capable of performing a wide array of tasks, are increasingly forming the foundation for numerous AI systems across the European Union. Recognizing that some of these models could pose systemic risks due to their capabilities or widespread use, the AI Act establishes specific rules for their providers, focusing on transparency and copyright.
To facilitate compliance and ensure a smooth transition into this new regulatory framework, the Commission rolled out the following three instruments in July 2025:
1. Guidelines on the scope of obligations for providers of GPAI models: These guidelines provide crucial clarity on who is responsible for complying with the GPAI obligations stipulated by the AI Act. They are designed to help various actors throughout the AI value chain understand their specific duties.
2. GPAI Code of Practice: Developed by independent experts and submitted to the Commission, this voluntary compliance tool offers practical guidance. It assists providers in adhering to their obligations under the AI Act concerning transparency, copyright, and critical safety and security aspects.
3. Template for the public summary of training content of GPAI models: This instrument mandates providers to offer a comprehensive overview of the data utilized to train their models. This includes detailing the sources from which the data was obtained, encompassing large datasets and top domain names. The template also requests information on data processing aspects, enabling parties with legitimate interests to exercise their rights under EU law.
These instruments are designed to operate synergistically, providing a clear and actionable framework for GPAI model providers to meet the requirements of the AI Act. The Commission emphasizes that this integrated approach aims to reduce administrative burden, stimulate innovation, and simultaneously safeguard fundamental rights and public trust in AI technologies.
Also Read:
- China Unveils Draft Measures to Enhance Ethical Governance in AI Development
- AI Governance Takes Center Stage: Industry Leaders Advocate for Balanced Policies to Maximize Societal Impact
The broader AI Act, which entered into force on August 1, 2024, represents the world’s first comprehensive legal framework on AI. It employs a risk-based approach, categorizing AI systems into four levels: unacceptable risk (banned practices), high-risk (subject to strict obligations), limited risk (requiring transparency), and minimal or no risk (largely unregulated). The governance and enforcement of the AI Act fall under the purview of the European AI Office and Member State authorities, with the AI Board, Scientific Panel, and Advisory Forum providing strategic guidance.
