On 8 November 2016, the International Atomic Energy Agency verified that Iran’s heavy-water stock had reached 130.1 metric tonnes. The agreement estimated needs at 130 tonnes at that stage and provided for excess material to be made available for export. The decimal place entered the record.^[1][2]

Iran subsequently reported that 11 tonnes had left the country on 19 November. On 6 December, the Agency verified the quantity at its destination abroad.^[1] Discovery, notification, shipment and independent confirmation were separate events. Keeping them separate makes it possible to see how a commitment was actually implemented, rather than merely announced.

The excess cannot be rounded away as close enough. Nor can a decimal place pronounce the end of an entire agreement. It gives the discrepancy an identifiable form. How serious it was, whether the response was adequate and what should follow still require reasons. The quantity does not contain its own complete judgment.

Diplomatic coverage often turns an agreement into a position to support or oppose. The figure 130.1 lets us start somewhere smaller. Something was detected, addressed and checked by another organization. Before choosing a verdict on the whole arrangement, we can ask what this sequence establishes—and what it cannot guarantee.

What a number can take out of the realm of promises

In 2015, Iran and China, France, Germany, Russia, the United Kingdom and the United States, with the European Union’s High Representative, agreed the Joint Comprehensive Plan of Action. The JCPOA combined specified nuclear measures, transparency arrangements and sanctions relief. Agreeing the text was not the same event as completing every action it envisaged.^[2]

Implementation Day followed on 16 January 2016, when the IAEA verified completion of the preparatory nuclear measures specified in Annex V.^[3] That finding had an object. It was not a declaration that Iran’s entire past was unproblematic, nor a permanent assurance about every future intention. A particular conclusion should not acquire a larger meaning merely because the agreement matters politically.

One strength of the arrangement was precisely its ability to make some questions smaller. Rather than relying entirely on assurances about intentions, participants could check specified stocks, equipment and records. Anxiety did not disappear. A portion of it could be addressed through repeatable observations instead of remaining indefinitely a dispute about what another government might secretly want.

The 130-tonne figure was not a boundary supplied by nature. Annex I, paragraph 14, set it out as an estimate of needs before the modernized Arak reactor entered service, with excess heavy water to be offered for export.^[2] The wording matters. An author cannot retain a number and then invent all the legal consequences that are supposed to accompany it.

That is why the November figure deserves to stay. Removing an inconvenient excess to demonstrate that verification worked would remove something verification actually did. Treating the same excess as immediate proof that all commitments had collapsed would omit the shipment and the subsequent check. Neither account would preserve the sequence it purported to explain.

This is not a search for a flattering name for a discrepancy. A record is useful because it prevents either side from selecting only the sentence it needs. There was a detected excess. There was also a documented response. An evaluation has to encounter both, rather than treating them as rival versions from which it may choose freely.

Verification consequently requires professional judgment: assessing information, checking it and deciding which conclusion it supports. That professional task has limits. Monitoring nuclear materials and specified commitments does not automatically make the Agency the judge of every military, diplomatic or regional activity of a state. Its expertise does not become unlimited merely because the political stakes are high.

Even a total has premises

Another episode complicates the image of a simple numerical test. A 2017 report addressed material retained in process equipment at Esfahan. A Joint Commission decision was made on 10 January; the associated treatment began later that month. Iran submitted its revised estimate of 99.9 kilograms of enriched uranium on 19 February, which the Agency found consistent with its assessment.^[1]

The number may appear substantial. But kilograms of uranium cannot be directly compared with a limit expressed in kilograms of uranium hexafluoride or its equivalent. The accounting treatment was also subject to conditions in the Commission’s decision.^[1] Precision about a number includes precision about what is being counted. A decimal place does not make its unit unnecessary.

Readers need not memorize those quantities. Their importance lies in the premises behind a total: which material counts, in which units and under which agreed conditions. A result can be numerically precise while still depending on a definition that needs explaining. Measurement does not eliminate interpretation; sometimes it makes a previously hidden interpretive question visible.

Public explanations matter for that reason. If people see only a final total without its accounting basis, supporters may read certainty into it while opponents read every clarification as a concealed concession. Accessible reasons will not remove disagreement. They can give disagreement something definite to address, rather than requiring it to revolve entirely around suspicion.

The agreement also provided a dispute-resolution route. Participants could refer a performance concern to the Joint Commission and pursue further procedures set out in the text. Whether an unresolved matter constituted significant non-performance was not a verdict automatically generated by a measuring device.^[2] A technical report was important without being the only institutional document that mattered.

This distinction should not become an escape route from evidence. Anyone calling a discrepancy serious must explain why it is serious. Anyone calling a correction sufficient must explain that too. The fact that a judgment remains to be made does not make established observations optional. Different responsibilities ought to make the account more demanding, not less.

The Agency’s May 2018 report marked an especially important boundary. It continued to verify non-diversion of declared nuclear material, while its evaluation of the absence of undeclared material and activities remained ongoing.^[4] Those statements did not contradict one another. A finding about declared material cannot simply be enlarged into proof that every possible undeclared activity has been ruled out.

The converse is equally necessary. An evaluation still under way is not itself a discovery of secret activity. If every incomplete conclusion is converted into incriminating evidence, honest qualification becomes indistinguishable from failure. The institution that acknowledges its limits is then punished for the very precision that makes its reports useful, while an unlimited assurance becomes easier to sell.

The reports continued; the decisions diverged

On 8 May 2018, President Trump announced that the United States would cease participating in the agreement and begin restoring relevant sanctions. His public reasons included the agreement’s duration, longer-term nuclear risks, missiles and regional security.^[5] These were the administration’s claims and policy assessment. They were not a technical judgment made on its behalf by the IAEA.

The British, French and German leaders reached a different conclusion that day. Referring to IAEA reporting, they considered Iran to be abiding by the agreement’s restrictions and argued for retaining it.^[6] The disagreement was not merely about whether the same column had been added correctly. It concerned whether a deliberately limited exchange remained worth maintaining.

On 24 May, a new IAEA report continued to list the results of monitoring, recording stocks and enrichment levels within the specified limits while preserving its qualifications about undeclared activities.^[4] A political rupture had not automatically converted the technical account into its opposite. The report still had to say what the Agency could substantiate.

A serious objection to retaining the agreement could ask whether functioning present restrictions were sufficient for longer-term risks and activities outside its scope. A serious argument for retaining it could ask what would remain after removing verified constraints, and how a better alternative could actually be obtained. Neither question is frivolous merely because it is not the other side’s preferred question.

A limited agreement does not establish that every relevant danger has been solved. Dangers beyond its scope do not automatically establish that restrictions inside it have no value. Expanding the purpose of cooperation requires an account of how new commitments can be agreed, rather than silently treating additional demands as if they had always been part of the original bargain.

Ordinary life lies at a further distance. Implementation Day relief concerned specified nuclear-related sanctions; it did not remove every sanction.^[7] Legal permission for some transactions is different from a bank’s willingness to process them, or a business’s ability to obtain finance. The path from an international announcement to an available service has more than one decision-maker along it.

The IMF’s contemporaneous 2017 study recorded restored banking relationships, but also the continuing absence of major banks. Remaining sanctions, enforcement risk, anti-money-laundering weaknesses and transparency concerns were among the obstacles.^[8] Relief cannot stand as proof of improvement for every household. Equally, obstructed transmission does not establish that no beneficial change occurred. Neither extreme follows from the available account.

The choice to continue or stop therefore cannot be explained only in the security vocabulary of negotiators. Assessments may differ, but they still need to address expected benefits, alternatives and costs borne by others. Verification cannot supply all those reasons. Its inability to supply them is not permission for decision-makers to stop offering reasons themselves.

This essay remains within 2015–2018. Later events cannot provide every participant with knowledge unavailable at the time. What the contemporary record already shows is verification operating, particular issues receiving concrete responses and participants disagreeing about the adequacy of the arrangement. Keeping that combination intact is more revealing than reading the whole period backwards from a later outcome.

Perhaps that is what the decimal place leaves us with. Evidence reduces what can be said arbitrarily without relieving anyone of the responsibility to judge. The figure has been entered. Who decides next, on what grounds, and at whose cost are still questions requiring an answer.