The European Union is spearheading a major initiative to develop the first "General-Purpose AI Code of Practice" under its AI Act.
Announced on September 30, this effort aims to set standards for transparency, risk management, and governance of AI models.
The European AI Office is leading the initiative, bringing together nearly 1,000 global experts from academia, industry, and civil society to shape the framework.
This new Code of Practice will focus on general-purpose AI models, such as large language models (LLMs) and AI systems integrated across various sectors.
It will become a key component of the EU’s AI Act, which was passed by the European Parliament in March 2024.
The act takes a risk-based approach to regulating AI, with different systems categorised into varying levels of risk, ranging from minimal to unacceptable.
The drafting process for the Code of Practice began with an online plenary session involving almost 1,000 participants.
Four working groups, led by prominent experts including AI researcher Nuria Oliver and copyright specialist Alexander Peukert, will focus on transparency, copyright issues, risk mitigation, and internal risk management.
These groups will collaborate over the next several months to draft provisions, gather feedback, and refine the code, with a final draft expected by April 2025.
According to the European AI Office, over 430 stakeholder submissions have already been received, reflecting widespread interest in shaping the future of AI regulation.
Despite concerns from major tech companies like Meta, which argue that strict regulations could hinder innovation, the EU is taking a balanced approach, aiming to foster innovation while ensuring ethical AI development.