Where the Miss Lands: Kuhestak, Minab and the Precautions that Come Before the Balance

[Neil Cameron is a barrister and an Affiliate Scholar at the Judith Reppy Institute for Peace and Conflict Studies, Cornell University.]

On the evening of 1 September a United States glide bomb came through the roof of a house in Kuhestak, a small port town on the Iranian shore of the Strait of Hormuz, where about 150 women and children were waiting for a bride and groom. The bombardment left five people dead and more than sixty wounded. United States officials have said the intended objective was a communications tower, used for both civilian and military traffic, just over a hundred metres away. Thirty kilometres inland, on 28 February, at least two United States missiles struck the Shajareh Tayyebeh primary school and an adjacent pharmacy in Minab. Airwars has identified 157 people killed, including 123 children, and places the eventual toll at between 157 and 168. The preliminary finding of the United States inquiry, as reported in March, was that the coordinates rested on outdated intelligence about an adjacent Revolutionary Guard naval facility.

The official response to both has the same shape: the United States does not target civilians; the incident is under review; mistakes happen. Asked about Kuhestak, the Vice President said that sometimes things happen. That moves in a single step from distinction to accident, and passes over the part of the law whose purpose is to prevent accidents. Calling civilian harm a mistake does not answer the prior legal question, which is whether feasible precautions were taken to avoid or minimise it.

The Obligation that Comes First

Article 57(2)(a)(ii) of Additional Protocol I requires those who plan or decide upon an attack to take all feasible precautions in the choice of means and methods of attack with a view to avoiding, and in any event to minimising, incidental civilian harm. Only after that, in sub-paragraph (iii), does the Protocol ask whether the harm that remains would be excessive in relation to the military advantage anticipated. That precautions are distinct from the excessiveness assessment is established; that they must be exhausted before the balance is reached now has substantial support: Dorsey has argued for the primacy of precautions as obligations to be discharged before the balance is struck, building on Van den Boogaard, for whom the excessiveness rule is a secondary rule that enters only where precautions cannot avoid the harm altogether (pp. 122-23), and Corn, who argues that all feasible precautions must be taken before any balance is struck (pp. 435-36). The narrower point pressed here, and argued at length elsewhere, is that the order is supplied by the text itself. “Avoiding” does work that “minimising” does not: it asks whether the harm can be removed by choosing differently, and that question is neither asked nor answered by weighing harm against advantage. A planner who goes straight to the balance has not complied with sub-paragraph (ii); the treaty does not permit the second question to absorb the first.

The United States is not a party to the Protocol, but it is bound all the same. Its Law of War Manual states that taking feasible precautions is a legal requirement (para. 5.2.3.3), and in 1987 the State Department’s Deputy Legal Adviser, declining the treaty, accepted the principle of all practicable precautions as one the United States supports and as reflected in Articles 57 through 60 (Matheson, 2 Am U J Int’l L & Pol’y 419, pp. 426-27). The precautionary obligation its own doctrine accepts is absent from the public account of both incidents.

Minab: Verification

Minab is in the first instance a failure of verification under Article 57(2)(a)(i). The school’s civilian character was not hidden, and its status was visible in open-source imagery. The deeper problem, as Woodcock and Dorsey have shown, is that the staleness of the underlying database was known: analysts responsible for the Modernised Integrated Database had warned for years that facility records were not being revalidated, and concerns about this school had reportedly been raised. Stale intelligence is not an act of God. It is a known failure mode of any targeting system built on a database, and the precaution against it, verification against current sources before authorisation, was feasible because the sources were open. AI-enabled decision support, as Shoker and Dorsey argue, compresses the time in which that verification can occur; it does not remove the obligation to perform it. Minab is also a question of method. Even on the United States’ own account of the coordinates, striking a naval facility beside a primary school at 10.23 on a school day was a choice of timing under sub-paragraph (ii), and the harm it produced was foreseeable.

Kuhestak: Means, Methods and the Direction of Error

Kuhestak is the clearer case for sub-paragraph (ii), because the analysis does not depend on contesting the objective: assume that the communications complex was a lawful military objective, and the harm was nevertheless produced entirely by how it was attacked. The precautionary inquiry is not whether the selected weapon could hit the tower. It is whether, among the means and methods feasibly available, another weapon, heading, aim-point, fuze, timing or release configuration could have materially reduced foreseeable civilian harm. The public record does not yet establish that any such alternative was feasible; what it establishes is that these are the legally relevant questions, and that the review must answer them.

Four stand out. Timing: a tower is a static object, and the record discloses no reason why it had to be struck at half past nine on a Tuesday evening rather than when fewer civilians were likely to be nearby. Weapon: the Post reports that six JSOW glide weapons were launched at the complex, the only target in the town, and that one veered off course; independent analysts have identified JSOW fragments at the scene. Why six weapons of that class were assigned to the communications complex, a hundred metres from housing, and what aim-points were selected and why, are questions the review should answer. Observation: Central Command is reviewing footage from drones that were monitoring the area; if a gathering of 150 people was detectable, the assessment had to be revisited and, if the revised expectation of civilian harm engaged it, the duty of suspension under Article 57(2)(b); if it was not detectable, the review should explain what the surveillance could show. Geometry: the United States’ own collateral damage methodology treats delivery heading as a way of mitigating both fragmentation and the risk created by delivery error, alongside aim-point offset and fuzing (Encl D, App A, Table D-A-1). Where an objective can be attacked from more than one direction, the comparison cannot stop at the probability of a hit; it must include what lies along the paths a delivery error would take. At Kuhestak the tower stood at the seaward edge of the town, with the harbour to its west, and the house lay inland, about 135 to 150 metres to the north-east. Whether another feasible heading would have displaced the consequences of error towards the sea is a question the United States’ own methodology recognises, and one the review is equipped to answer.

Feasibility and Risk

The obvious reply is that a different heading, a lower delivery profile, a longer loiter or a delayed strike might have exposed the aircraft to greater risk. Force protection is a military consideration that bears on feasibility; but a precaution is not infeasible merely because it costs something, or sub-paragraph (ii) would have almost nothing left to require. Walzer’s due care argument makes the moral point: soldiers must accept some risk to themselves in order to reduce the risk they impose on civilians, though not without limit (pp. 155-56 and later). Feasibility is also dynamic. As Dorsey argues in her working paper on precautions-by-design, technological and informational capability can widen the precautionary space as readily as narrow it, and feasibility cannot be narrowed by constraints an attacker has created for itself; it would be a strange law under which a military’s targeting capability grew while the precautions it deemed feasible shrank.

That is why demonstrated capability matters, as evidence. Feasibility is not identical to technical capacity; it is what was practicable in the circumstances, including military ones. But feasibility is indexed to capacity, in the sense that what an attacker has demonstrated it can see, verify and adjust is the best available evidence of what was practicable for it (what I have elsewhere called capacity indexing), and the more it has shown it can do, the heavier the explanatory burden, not a burden of proof, on the party holding the targeting records when it did not. On these two incidents the United States has shown a great deal: drones over Kuhestak that night; open-source imagery of the Minab school; a methodology that quantifies stray probability and prescribes heading and fuzing as mitigations; and, until this year, a civilian harm mitigation apparatus whose no-strike lists and pattern-of-life work Central Command itself described as operationally effective. That office has since been largely dismantled, which bears on the institutional context in which Kuhestak occurred and on what the review can be expected to produce. Wes Bryant, a former targeter who later served in that office, told the Post that pattern of life should be established and every adjacent structure characterised before any attack in a populated area. The United States has described the Kuhestak targeting process as exhaustive; if so, the records that would answer these questions exist.

What is Missing

Heading, aim-point, timing and weapon selection appear in most Western manuals as ways to reduce or minimise incidental harm. What the manuals do not say is that these choices must be concluded before the balance rather than inside it, and “reduce” cannot carry a rule of priority. The United States’ own methodology, the working document of the quantification logics Dorsey has traced, is closer to the treaty than its manual: of the five basic questions it requires before any engagement, whether harm can be mitigated by a different weapon or method comes third, and whether the remaining harm would be excessive comes fifth. The order is present in the working document and absent from the doctrine that is supposed to explain it, which is a defect of commentary and diagram rather than rule text, and can be corrected without any state changing its legal position.

Two incidents in the same province, six months apart, explained the same way. Whether the harm was a mistake is the question the reviews have been asked. Whether it was avoidable is the question the law asks first.