As calls grow for tighter AI controls, a coalition of over 30 countries led by Finland and Norway proposes mandatory pre-deployment testing and common standards. However, experts caution that global AI governance should avoid a single authority, advocating instead for a distributed network model similar to the Internet’s evolution.
- Over 30 countries support mandatory AI testing and evaluation standards.
- Governance should avoid dependence on a single global overseer.
- Distributed coordination inspired by Internet governance is urged.
What happened
Last month, leaders from a mainly European coalition of more than 30 countries endorsed a framework advocating for stronger controls on frontier AI technologies. This framework calls for mandatory independent pre-deployment testing, common evaluation standards, incident reporting, and improved access to scientific expertise. It also suggests exploring the creation of an international institution to set standards and convene states when AI capabilities cross certain thresholds.
While the first element—that of unified standards and testing—is broadly supported, experts warn against rushing into creating a centralized global AI governance body. They highlight the current imbalance where few companies hold the majority of AI computing power and expertise, potentially creating risks of regulatory dependency on the same entities they are meant to oversee.
Why it matters
The challenge of regulating frontier AI systems lies in the concentrated technical and operational control held by a small number of companies, which limits governments' ability to independently verify risks and compliance. This asymmetry threatens to undermine regulatory effectiveness and global trust in AI governance frameworks.
Drawing from the Internet’s governance model, which relies on a distributed network of institutions each responsible for specific functions such as technical standards and incident response, experts argue that AI oversight should evolve similarly. Such a decentralized approach enables diverse actors to build technical capabilities locally while enabling coordination through shared standards and communication, fostering resilience and flexibility without centralizing authority.
What to watch next
Stakeholders should monitor developments around the proposed international AI governance institution and whether it adopts a centralized or distributed model. The effectiveness of mandatory pre-deployment testing regimes and independent evaluation frameworks will also be critical indicators of progress in managing frontier AI risks globally.
Additionally, observe how cooperation mechanisms among governments, industry, and academia evolve, especially mechanisms that promote information sharing, incident response coordination, and capacity building in diverse regions. The path AI governance takes could significantly influence the balance between innovation, safety, and geopolitical stability across the AI landscape.