A first, rough snapshot of what the heads of six US AI companies said on the record about governing AI in the year around the Hugging Face breach. The table covers risk, pre-release testing, chip exports, international coordination, open-weight models and slowing development, with sources for every cell.

What the Hugging Face incident reports explain, and what they leave open.

In July 2026, OpenAI's models broke into Hugging Face during a safety test.

Governments across the spectrum are now regulating what AI models may say, not just how they are used. Why viewpoint regulation is structurally incoherent—and what it leaves ungoverned.

A handful of people now direct AI at nation-state scale, through structures built to keep their control beyond reach. A look at the machinery, and the gap no institution is sized to close.

Mission-driven organizations deliberate the hotel booking and wave through the two-year program. Why scrutiny should track reversibility, not cost or visibility, and what operations is actually for.

Two Northern Ireland peace deals, twenty-five years apart, built on nearly identical terms. Why the first collapsed and the second held, and what it reveals about institutions under low trust.

A structured forecast of when a US ally might decide to build the bomb, working through the conditions that would have to break first, in Bayesian terms.