Middle Powers Can Lead on AI Safety

In this commentary piece for Project Syndicate, authors Gordon LaForge & Robert C. Orr talk about the growing risks associated with frontier AI development and the lack of meaningful regulatory action in the United States. Against a backdrop of increasingly serious warnings about the potential dangers of superhuman AI, including concerns over loss of control and inadequate safety practices, they argue that voluntary industry commitments are insufficient to address the challenges ahead.

The authors suggest that middle powers could play a pivotal role in establishing international safeguards. More than 30 political leaders have already endorsed a call for greater coordination and the creation of an international AI agency. Countries such as India, Japan, South Korea, Brazil, South Africa, and EU member states could use their collective economic and diplomatic influence to push AI companies toward common safety standards.

One particularly effective approach would be to make access to their combined markets conditional on meeting agreed safety requirements, including mandatory pre-deployment testing and independent audits. Governments could also pool resources to strengthen AI safety research institutions, share technical testing facilities, and establish common incident-monitoring and reporting systems to identify emerging risks.

Looking ahead to the Geneva AI Summit in June 2027, LaForge and Orr argue that middle powers should develop a coordinated safety agenda and establish clear international red lines for frontier AI development. While securing agreement will be difficult, they point to precedents in nuclear safety, pandemic preparedness, and climate policy to demonstrate the potential of collective action. Ultimately, broader international cooperation will be essential to ensure AI development remains safe and beneficial, while encouraging the United States and China to adopt stronger safeguards…

CONTINUE READING…