The EU is assembling experts to develop the first Code of Practice for AI models under the AI Act, setting new guidelines for risk assessment and transparency.
With the creation of the first “General-Purpose AI Code of Practice” for AI models under the AI Act, the European Union is taking a significant step toward influencing the direction of artificial intelligence.
The European AI Office is leading the initiative, which brings together hundreds of international experts from academia, business, and civil society to collaboratively draft a framework that will address critical issues like transparency, copyright, risk assessment, and internal governance, according to an announcement made on September 30.
Almost 1,000 people are influencing the future of AI in the EU.
With almost a thousand attendees, the kick-off plenary online signaled the start of a months-long process that will culminate in the final draft in April 2025.
Applying the AI Act to general-purpose AI models, such as large language models (LLMs) and AI systems integrated across several industries, is expected to be based on the Code of Practice.
In addition, four working groups headed by eminent industry chairs and vice-chairs were announced during this session, and they would be responsible for developing the Code of Practice.
Well-known professionals like German copyright law specialist Alexander Peukert and artificial intelligence researcher Nuria Oliver are among them. These working groups will concentrate on internal risk management, technical risk mitigation, risk detection, transparency, and copyright.
The European AI Office reports that these working groups will convene from October 2024 to April 2025 to develop clauses, obtain feedback from interested parties, and continue to discuss improving the Code of Practice.
Preparing the ground for international AI governance
The European Parliament passed the EU’s AI Act in March 2024, a significant legislation that aims to regulate technology throughout the EU.
It was developed to provide a risk-based framework for the governance of AI. It assigns multiple risk categories to systems, from negligible to intolerable, and requires particular compliance actions.
Because of their wide range of uses and potential to have a significant social impact, general-purpose AI models are particularly pertinent to the act and frequently fall into the higher-risk categories described by the law.
However, a few significant AI firms, including Meta, have objected to the rules, claiming they are overly demanding and may impede innovation. In response, the EU worked cooperatively to establish the Code of Practice to balance ethics and safety while promoting innovation.
More than 430 contributions have been made to the multi-stakeholder consultation, which will impact the code’s drafting.
The EU hopes that by April of the following year, the results of these efforts will provide a standard for the responsible development, application, and management of general-purpose AI models, focusing on reducing risks and optimizing benefits to society.
This endeavor will probably impact AI regulations globally as the field of AI develops quickly, particularly as more nations turn to the EU for advice on how to regulate cutting-edge technologies.
