This paper documents four structurally similar episodes from June 2026 in which different nodes of the U.S. government reached opposite or uncoordinated conclusions about the same AI company within the same month: California's statewide partnership with Anthropic running alongside the Department of War's unresolved "supply-chain risk" designation; a multi-agency standoff over the export-control suspension of Fable 5 and Mythos 5; an intra-department reversal on the CXMT/YMTC entity list; and a non-transparent comparative application of EO 14409 to Anthropic versus OpenAI. The paper does not argue conspiracy or incompetence. It documents a structural pattern: U.S. AI policy in June 2026 was produced by at least five federal nodes acting without a binding coordination mechanism, alongside a sixth, independent state-level track. The pattern is falsifiable; its first verification checkpoint is the Genesis Mission Initial Operational Capability date of August 21, 2026.
Marco Perugini (2026) studied this question.