ASI 2030's records on international order, all of them synthetic scenarios, make one claim through their opinions: a judgment about another side can be audited only as far as the evidence behind it can be seen from outside the side that holds it. This paper has followed that claim along the studio's line, from the shared past through the future with an accord and the future without one.
Discussion
The three settings give the auditors three problems. Before the fork it is inference: public facts about the other party are read for intention, and the auditors accept the facts while refusing the step, naming what it needed, a private channel, a stated position on the harder clauses, cases from other regions, an adviser shown to be independent. In the future with an accord it is declaration: a side's evidence about itself arrives as that side reports it, and where a judgment rests on such a report the auditors give no opinion, while a judgment resting on the text alone can be audited and exposes a gap in what the text covers. In the future without an accord it is identity: the party that reports and the party that checks become one, and the outcome splits on whether an outsider's own result can be set against the judgment.
Set beside the debates, the records give each a concrete turn. Chayes and Chayes (1993) treat the general level of compliance as unverifiable and give reasons for assuming a propensity to comply; the records' auditors cannot choose an assumption for one judgment, and their disclaimers are where unverifiability becomes an outcome on the page. Downs, Rocke and Barsoom (1996) warn against reading easy compliance as good news about hard cooperation; the secretariat's memo makes that inference and draws the trunk's one adverse opinion. Abbott and colleagues (2000) separate obligation, precision and delegation; the records keep showing delegation as the missing dimension, from an accord whose interpreter has not yet spoken to a branch in which no third party remains to measure. Fearon (1995) and Jervis (1978) place private information and the legibility of posture at the root of failed cooperation; the records turn these into questions an auditor can ask of one judgment. The proposals of Ho and colleagues (2023) and of Trager and colleagues (2023) imagine bodies that would monitor or certify; the records imagine such bodies' absence, and an accord's partial presence, case by case.
Contributions
The paper makes three contributions. First, it shows the opinion type working, in this synthetic record set, as a measure of mutual access: qualified where signals are read without the channel that would confirm them, a disclaimer where the evidence is a side's account of itself, adverse, in the future without an accord, where an outsider's own result contradicts the judgment. Second, it identifies the auditors' distinction between kinds of agreement, accepting a result as against agreeing to disclose and cross-audit, and reads it against the debate on compliance and on legalization, where the difference between easy and hard commitments is central. Third, it shows that in the future without an accord the reopening of a judgment is tied to an institution: every reopening line there waits on a body or a channel, so that revisability itself depends on a third party existing. The paper does not judge which future is better, which opinion any auditor should have given, or what any real delegation should do.
Limits
The evidence is eleven records, all synthetic, written by one studio to a single form. They report nothing about real states, treaties or AI systems, and the paper's findings are about how this record set frames decisions, not about international relations. The counts come from keyword rules over short texts, and their categories, such as a party's report of itself or an outside result on file, are rough. The frame was measured only within the eleven records, where no five-word run recurred in three or more; the studio's full library was not measured, so phrasing shared with records on other axes may have gone unseen. The works are read through short reading notes, not in full, and several fit the records only partly: none of them is about audit, and the security works concern war, which the records do not imagine. The records were read in English as the studio published them, and the paper treats the order of the studio's line as given.
Further research
Three lines of work follow. One would read the other axes of the same library for the same pattern, asking whether the link between the opinion and the source of evidence holds where no border is crossed. A second would set the synthetic records beside the design documents of real international verification arrangements, not to take facts from the scenarios but to ask whether the questions the auditors put are the ones such arrangements are built to answer. A third would follow the reopening lines over time: if the studio publishes later records in which a measurement body comes to exist or a data channel reopens, those records would show whether the judgments waiting on them are in fact taken up again.