Man-made power outage episode shows that frontier AI requires global governance

A U.S. government directive requiring Anthropic to suspend foreign access to its AI models Fable 5 and Mythos 5 sparked a broader debate about the development and deployment of artificial intelligence and raised concerns around the world. Calls for technological sovereignty are growing louder, with public concerns about the impact of a single government being able to restrict access to cutting-edge AI systems without consultation or transparency. Equally pressing concerns are concerns about the risks associated with implementing such frontier models in the absence of independent oversight and a strong global governance framework.

Calls for a strong and reliable regulatory framework are also growing, with even industry leaders advocating for increased oversight. A few days ago, Anthropic CEO Dario Amodei argued that “the time has come to move beyond transparency to deeper and more binding AI regulation.” While the current issue reportedly stems from concerns about circumvention of existing safety measures, the development of this decision raises deeper questions. As Anthropic itself noted, the directive “does not provide specific details of national security concerns.” The fundamental question is: Who decides access to frontier AI models? Should such decisions be left to governments, which have traditionally used chip export controls to restrict access to critical AI hardware, or to private company executives, whose decisions may be shaped by commercial and strategic considerations? What is needed is a transparent framework built on clearly defined principles and procedures. In fact, even Amodei argues that Frontier models “should be required to undergo technical testing and auditing, and failure to meet high safety standards should result in their release being blocked or withdrawn as a threat to public safety.” He recommended that testing be carried out “by a qualified third party.”

Given the transformative and borderless nature of AI, these complex problems cannot be solved by single countries acting alone. These need to be negotiated at a global level through a governance architecture built on broad consensus and inclusive participation. All stakeholders must have a seat at the table. With one of the world’s largest AI user bases and vast amounts of data that contributes to the value and functionality of global AI systems, India is a key stakeholder in this debate. We must play an active role in shaping the international regulatory framework governing the future of AI.