Nicolas Moës / Nov 2024
Image: Shutterstock
An important process is currently underway in Brussels. Led by the European AI Office and a team of expert Chairs and Vice-Chairs, major international AI companies, industry, civil society and academics are coming together to draft the Code of Practice (Code) governing general-purpose AI (GPAI), i.e. the most advanced and widely-applicable AI models. The Code is crucial for the implementation of the groundbreaking EU AI Act provisions applicable to GPAI model providers as of 2 August 2025; it presents the first opportunity to detail the rules for placing GPAI products and services on the EU market.
But, for some time there has been speculation about whether the EU AI Act could spark a “Brussels Effect” of increased AI governance beyond the EU Member States. Now, as the multi-stakeholder Code drafting process accelerates towards the first working group sessions in mid-November, the possibility that the Code could be a key tool for activating a Brussels Effect outside Europe is becoming a reality.
Why does the Code of Practice matter to Europe?
First, the Code can induce an ecosystem-wide shift towards greater transparency and understanding about GPAI models. This will benefit regulatory authorities, rightsholders and key stakeholders along the value chain as they gain access to key information about GPAI models that are otherwise opaque. For example, for smaller European companies who find themselves in the downstream AI value chain, greater transparency can help to re-balance commercial information asymmetries. Those asymmetries impede smaller companies’ bargaining positions and compliance with their own obligations under the EU AI Act.
Second, the Code can carefully marry legal certainty with flexibility, enabling the EU AI Act’s regulatory regime to respond to the ever-evolving technology. Legal certainty offers current and future AI businesses clarity about what is (not) permissible, a delineation which, alone, could stimulate entrepreneurship and investment in the AI ecosystem. Giving GPAI model providers guidance to invest confidently and responsibly in safety and other governance procedures for their models is to be encouraged. Flexibility renders the EU’s AI regulatory regime future-proof, as the development of GPAI models, and how they are evaluated, advances. Offering GPAI model providers some latitude in evidencing precisely how they satisfy their obligations has the particular advantage of encouraging new and improved evaluation methods to emerge that can be embraced by the Code’s requirements.
Third, although controversial, the Code can be testament to the fact that the oft-repeated regulation versus innovation dichotomy is false. As the arguments above highlight, the Code can be a catalyst for responsible AI innovation. Once the Code states actions that GPAI model providers are to perform, or the outcomes they are to achieve, for compliance with the EU AI Act, consumers will gain confidence in GPAI models and systems that they know they can trust and are willing to welcome into their professional and personal lives. Europe should be proud to foster AI innovation built on a consumer demand, and an entrepreneurial respect, for trustworthy and responsible AI.
Why does the Code of Practice matter to the rest of the world?
First, the Code can establish a responsible and trustworthy approach to GPAI model development in the AI companies outside Europe that are placing products on the European market. The vast majority of the most influential GPAI model providers are located in the United States or the United Kingdom. Neither of those jurisdictions has its own comprehensive regulatory regime. Therefore, the potential for a governance gap to be filled by spreading a European-compliant approach is clear. Europe’s first-mover advantage, and its sizable consumer base, can be leveraged through the Code.
Second, the Code can be an exemplar for translating legislative obligations into action-oriented measures and key performance indicators that will become ingrained in the day-to-day practices of designing and developing GPAI models. By demonstrating to policy-makers across continents the actions necessary to make robust governance part of daily business life, the Code can trigger behaviour change in the development of GPAI models that will benefit consumers worldwide.
Third, the Code can be a global blueprint for a genuinely collaborative process of co-regulation. By harnessing the knowledge of industry leaders and balancing that with non-industry expertise and a diversity of counterpoints, regulatory capture can be avoided. This is essential for the Code and its equivalents in other jurisdictions to be the credible, responsible and trustworthy instruments of governance that we all need and deserve.
In short, Europe–and the world–is at a turning point. As policy and legal minds now focus on what the Code will stipulate as compliance with the EU AI Act, there is a growing recognition that the Code matters to the rest of the world. As a major piece of legislation on AI with global significance, the EU AI Act is charting a new path forward for AI governance and innovation. The GPAI Code of Practice can showcase how intention becomes action.
