The Moment
Frontier AI is already global. Its most consequential capabilities are advancing toward risks that could be catastrophic, irreversible, and impossible to contain within national borders. Yet the institutions ultimately responsible for governing it remain territorial, politically cyclical, and dependent on continued coordination among competing states. The International AI Safety Report 2026 documents cross-border risks and uncertainty about their development.
Who is responsible when a system developed in one place creates consequences somewhere else? That question connects the technical work of evaluating AI with the human work of deciding who can authorize it, challenge it, or restrict its use.
SIRDA focuses on AI that contributes to developing more capable AI. Recursive self-improvement describes a feedback process in which improvements increase a system's ability to produce further improvements. The inquiry examines how that process changes oversight, including when systems gain access to laboratories, infrastructure, and robotics. Its pace, limits, and feasibility remain questions for evidence.
The public conversation is widening too. On September 15, 2026, the Center for Humane Technology announced the Netflix release of The AI Doc: Or How I Became an Apocaloptimist. Its account describes director Daniel Roher exploring what AI means for the future as he prepares to become a parent. That brings a human question into view: who gets a meaningful say in decisions whose consequences reach beyond the people making them?
The week of September 21, 2026 showed how quickly the question is moving. Finland and Norway launched a Call for Control of Frontier AI Models, backed at launch by 22 leaders from 20 countries and the European Commission, asking UN member states to explore creating an international institution able to set standards, enable verification, and convene states when capability thresholds are crossed. The United States, China, and the United Kingdom are not among its endorsers. Two days later, the UN Security Council held a high-level briefing on frontier AI and international security. Then Washington and Beijing established a dialogue on advanced AI and agreed to create a channel for AI incidents. Each step matters. None yet specifies who decides, on what evidence, when informed parties disagree about a dangerous development. That is the gap this inquiry examines.