Analysis
Analytical notes on AI governance events, institutional signals, and structural developments as they occur.
These pieces are not part of the Beyond Control series. They apply the series’ analytical framework to specific documents, decisions, and moments that warrant structural examination.
Second Table, Same Ceiling: Two Systems of Legitimation, One Structural Limit
On July 16, 2026, twenty-nine states signed the founding agreement of the World Artificial Intelligence Cooperation Organization (WAICO), Uzbekistan among them. Three weeks earlier, the United States and some three dozen partner economies had signed the AI Opportunity Statement at the second Pax Silica Summit. Within a single month, global AI governance acquired a second institutional table, built on the opposite logic of membership. Yet the two registries barely overlap, and neither carries the one capacity the first was already found to lack: the ability to independently verify the systems both exist to govern. More tables, the same ceiling.
July 19, 2026. Read Analysis →
Russian-language version available
The Access List
On June 12, 2026, an export control directive suspended access to two frontier models for any foreign national, inside the United States or outside it, including the developer’s own employees. Ten days earlier, Executive Order 14409 had established a voluntary pre-release framework and expressly disclaimed the creation of any mandatory licensing or preclearance requirement. Two weeks later, GPT-5.6 was withheld from broad release pending federal review, with the government participating in the selection of who could use it. Nothing was contravened. The operative criterion was not where a system was deployed, but who was using it — and for institutions whose operations depend on capability licensed elsewhere, that distinction is the whole of the risk.
July 13, 2026. Read Analysis →
Russian-language version available
The Verification Ceiling: What Geneva Confirmed About Global AI Governance
On July 1, 2026, the UN’s Independent International Scientific Panel on AI released its preliminary report, naming an “evidence dilemma” at the heart of global AI governance. Five days later, the first UN Global Dialogue on AI Governance convened in Geneva and produced no enforcement mechanism — not because it failed, but because verification capacity, compute concentration, and a pre-declared sovereign limit made none structurally available. Three structural constraints, confirmed on the record, in the words of the institutions’ own participants.
July 9, 2026. Read Analysis →
Russian-language version available
From Incident to Institution: the Presidential Memorandum NSPM-11 and the Permanent Architecture of Sovereign Override
On June 5, 2026, President Trump signed National Security Presidential Memorandum 11 (NSPM-11), “Artificial Intelligence in the National Security Enterprise” — formalising a standing government right to compel continued AI system access and terminate any supplier that restricts how its technology is used. The clause is not abstract doctrine. It is the direct institutionalisation of the sovereign override first tested, in contested form, three months earlier in the Anthropic-Pentagon dispute. What a contract dispute predicted in March, a memorandum made permanent in June.
June 30, 2026. Read Analysis →
Russian-language version available
A Pause You Cannot Enforce
On June 4, 2026, Anthropic published ‘When AI Builds Itself’ — disclosing that most of its production code is now written by its own model and proposing a globally coordinated pause on frontier AI development. The proposal diagnoses the trajectory with precision. It does not explain why the proposed remedy is structurally unavailable, or what the pause debate means for the nations that will be bound by a decision made in their absence. An open letter to the consumer nations of artificial intelligence.
June 18, 2026. Read Analysis →
Russian-language version available
When the Center Speaks for the Periphery: AI Governance and the Limits of Moral Declaration
On May 15, 2026, Pope Leo XIV released Magnifica Humanitas — the first encyclical in the tradition of Catholic social teaching to address artificial intelligence directly. This is not a theological event. It is an institutional intervention in a contest over normative authority, arriving at the precise moment when that contest can still be shifted before infrastructure decisions harden into irreversibility. A structural reading from the periphery.
June 2, 2026. Read Analysis →
Russian-language version available
Subscribe to receive new work: okhodjaev.substack.com
Oybek Khodjaev researches AI governance and advises on institutional readiness in the age of AI, with a particular focus on systemic risk. Drawing on more than 35 years of experience across banking, finance, public administration, and business in Uzbekistan and the CIS, he develops practical approaches to assessing the institutional limits of AI governance. He is the author of the “Beyond Control” essay series, which advances a structural theory of the limits of AI governance.