1.2 asked what part each actor plays in a declaration: who owes one, who holds evidence about one, who checks one, and who no declaration covers. That map is a description. This is what you do with it.
Verification is not a property an actor has. It is a relation between two of them: somebody can put a fact in front of a verifier about somebody else, and that fact settles one of the four things a verifier has to establish. Drawing those relations is the exercise below, and the key for it is Baker et al.'s — the same report module 2.1 assigns when it gets to hardware.
Two things worth knowing before you start. Actors with no edge at all are a real answer and there are more of them than you would guess. And the point of counting is not the score: it is which of the four subgoals turns out to be holding the whole regime up.
Notes and Sources
The framework, the four subgoals and every mechanism quoted in the key: M. Baker, G. Kulp, O. Marks, M. Brundage & L. Heim, “Verifying International Agreements on AI: Six Layers of Verification for Rules on Large-Scale AI Development and Deployment” (2025). Quotations are matched against the committed copy of the paper at test time, so a quote that drifts fails the build rather than sitting on the page. Module 2.1 assigns the same report for its layers; this section uses its subgoals.
The cast, the rings and the sentences that place each actor on one are 1.2's, and the workshop above opens on the board you built there.