Estimated reading time: 15 minutes
In post #19, I asked how a commander reaches a sound assessment of a target’s status under uncertainty. This post asks what the law does when his information runs out. The two rules commonly discussed under the rubric of the presumption of civilian status in case of doubt are set out in Articles 50 (1) and 52 (3) of Additional Protocol I to the Geneva Conventions of 1949 (AP I) and addressed in operational doctrine across NATO and beyond.
The two provisions are doctrinally distinct but operationally adjacent. Article 50 (1) AP I directs that in case of doubt, a person be considered a civilian. Article 52 (3) AP I directs that an object normally dedicated to civilian purposes be presumed not used for military purposes. The verbs differ. The triggers differ. The scopes differ. Both rules answer the same operational question. What does the law require when doubt survives a reasonably executed inquiry?
The U.S. Department of Defense’s 2023 revision to its Law of War Manual brought this question back into practitioner focus. The revision adopted a “presume” formulation that, on close reading, goes beyond what AP I requires. This post traces why and proposes the corrective reading. It closes by mapping the doctrinal convergence with the reasonable-commander standard developed in post #19.
My substantive position is straightforward. Neither Article 50 (1) nor Article 52 (3) AP I requires the elimination of all doubt to protect civilians and civilian objects from attack. The rules require a reasonably executed verification process and a reasonable assessment in the circumstances. Residual doubt does not, of itself, translate automatically into a duty to abstain.
I. The text and its deliberate asymmetry
Articles 50 (1) and 52 (3) AP I share a function: each tells the commander what to do when the status or use of a target cannot be conclusively resolved. They do not share a form. The two articles use materially different verbs and address materially different subjects. Reading them as a single doubt rule with two objects obscures what the drafters wrote.
Two rules, two verbs
Article 50 (1) AP I provides that in case of doubt whether a person is a civilian, that person “shall be considered to be a civilian.” Article 52 (3) AP I provides that in case of doubt whether an object normally dedicated to civilian purposes is being used to make an effective contribution to military action, it “shall be presumed not to be so used.” The verbs are different and the difference is deliberate. At the 1974–1977 Diplomatic Conference, the Working Group on Article 50 agreed that ‘the concept of presumption gave rise to such difficulties’ that prompted the substitution of ‘considered’ for ‘presumed’ in the persons rule (CDDH/50/Rev.1, at p. 239, para. 39). The stronger ‘presumed’ verb was retained only for the narrower class of objects in Article 52 (3).
That choice has consequences for what the rules actually require of the commander’s reasoning, which Section III examines. At the textual stage the point is simpler: the drafters considered the question of which verb to use, made different choices for persons and for objects, and recorded their reasons. Throughout this post I use “presumption of civilian status” as the established term of art in IHL, U.S. and U.K. doctrine, and tribunal jurisprudence, but the textual asymmetry remains live below the shorthand.
Two scopes, not one
The scopes differ as visibly as the verbs. Article 50 (1) AP I addresses any person whose civilian status is in doubt. Article 52 (3) AP I addresses only objects “normally dedicated to civilian purposes” — a defined subset that the article itself illustrates: “a place of worship, a house or other dwelling or a school.” The rule for objects is triggered only when doubt arises about whether such an object is being used to make an effective contribution to military action. On its terms, the rule does not extend to objects whose nature, location, or purpose is in doubt — only to whether a normally-civilian object is being used militarily.
The asymmetry of scope is therefore the second deliberate feature of the AP I drafting. Article 50 (1) is a rule about status. Article 52 (3) is a rule about use of a defined class of object. My reading is that the drafters did not write a single doubt rule with two subjects; they wrote two rules with related but distinct triggers, distinct verbs, and distinct fields of application.
II. The trigger: what counts as doubt
Both rules engage on “doubt”. Additional Protocol I does not define the word, and state practice has not settled on a verbal threshold. What the practice does settle is something narrower but operationally more important: the rules are not engaged by every theoretical doubt a commander could entertain. They are engaged by doubt that survives a reasonable inquiry.
A divergent practice with a convergent core
State practice on the verbal threshold of doubt is honestly divergent. As Mina Radončić and Ashley Stanley-Ryan note, the formulations range from “benefit of the doubt” through “substantial doubt” to weighing the degree of doubt against the consequences of not attacking. The UK position, recorded both on ratification of AP I and in the UK Manual JSP 383, gives the most operational formulation for persons: it is “only in cases of substantial doubt, after this assessment about the status of the individual in question” (JSP 383, at para. 5.3.4) that the rule applies, and the rule does not override the commander’s duty to protect troops or preserve the military situation. For objects, the UK Manual tracks Article 52 (3) AP I directly: in cases of doubt, objects normally used for civilian purposes are “presumed as not being used for military purposes” (UK Manual, at para. 5.4.2).
These verbal formulations do not converge. The operative test underneath them does. The ICTY Trial Chamber in Galić framed the doubt rule for persons in terms a commander can apply: “a person shall not be made the object of attack when it is not reasonable to believe” that the target is a combatant, judged in the circumstances of the person contemplating the attack and on the information then available to him (Galić Trial Judgment, at para. 50). The ICRC Interpretive Guidance on Direct Participation in Hostilities reaches the same point from the other direction: the standard “must reflect the level of certainty that can reasonably be achieved in the circumstances” (Interpretive Guidance, at p. 76), with the intelligence available, the urgency of the situation, and the risks of an erroneous decision among the factors to be weighed. Read together, these sources point to reasonableness assessed in the circumstances of the attack, not against a fixed numerical threshold of certainty. The verbal labels diverge; the operative test does not.
No requirement of zero doubt
The doubt rules do not stand alone in AP I. They presuppose Article 57 (2) (a) (i) AP I — the obligation on those who plan or decide upon an attack to “do everything feasible to verify that the objectives to be attacked are neither civilians, nor civilian objects” and are military objectives within the meaning of Article 52 (2) AP I. This is the precautions duty that runs ahead of the doubt rules’ engagement. The commander has an antecedent obligation to take all feasible steps to dispel doubt before doubt becomes the trigger for the protective rules.
My reading is straightforward. The doubt rules engage on unresolved doubt — doubt that survives a reasonably executed verification. They do not engage on doubt the commander could have dispelled, in which case the persisting doubt is evidence that the precautions duty has failed rather than a trigger for the protective rules. This is the doctrinal foundation for the position that the doubt rules do not require zero doubt: they cannot, because they presuppose a process that has already reasonably narrowed it.
The 1987 ICRC Commentary on Article 50 AP I (para. 1920) records that persons of doubtful status “should be considered to be civilians until further information is available” — and adds that they “should therefore not be attacked” — a gloss that should be read with care. The Commentary’s tail flattens AP I’s “considered” into a duty to abstain pending verification, which is the same move the 2023 DoD Law of War Manual makes through different vocabulary, and one that the preceding analysis and Section III show AP I does not require.
The next section turns to what the rules set forth in Articles 50 (1) and 52 (3) AP I then require, when doubt does survive that reasonable inquiry — and to the doctrinal puzzle that the 2023 revision to the U.S. Department of Defense Law of War Manual has placed at the centre of contemporary practice.
III. The test: what the rules require of the commander’s process
The rules’ existence is no longer ignored; it is settled. What matters now is what the rules do to the commander’s reasoning. This Section will first show why the 2023 revision to the U.S. Department of Defense Law of War Manual (“DoD Manual”) went too far and then what is and isn´t really required to engage a target legally in the fog of war.
A. The 2023 reversal
Before July 2023, § 5.5.3.2 of the DoD Manual stated that “[u]nder customary international law, no legal presumption of civilian status exists for persons or objects.” The 2023 revision relocated the discussion to § 5.4.3.2 and reversed the substantive position: commanders “must presume that persons or objects are protected from being made the object of attack unless the information available at the time indicates that the persons or objects are military objectives.” The companion provision at § 5.4.3.4 adds that the § 5.4.3.2 discussion “reflects the DoD view of customary international law applicable to assessing whether persons or objects are military objectives, including in cases of doubt.” James Schoettler characterises this as a complete reversal of the DoD’s prior position on the customary status of the rules.
The 2023 revision settled the U.S. position but not the broader customary-law question. The U.S. has adopted a particular view of what is customary — a single “presume” formulation that flattens Articles 50 (1) and 52 (3) AP I into one rule. As Schoettler observes in his 2024 analysis, the ICRC’s 2005 Study of Customary International Humanitarian Law did not adopt either AP I doubt rule as a black-letter customary rule, and he concludes that the § 5.4.3.4 statement is best read as “the definitive U.S. position on the CIL status” rather than as a settled international consensus. The 174 States Party to AP I remain bound by the treaty’s two-rule, two-verb formulation. What state practice does support is something narrower than either the U.S. formulation or the AP I treaty rules: the common principle that doubt about the status of persons or objects must be given due consideration before engaging a target.
That a single revision could carry the U.S. from “no legal presumption exists” to “the rule reflects customary international law” deserves explanation. AP I does not require what the 2023 Manual now requires.
B. Why the 2023 DoD Manual goes beyond AP I
My reading is that four converging causes explain the over-correction.
First, the textual asymmetry of AP I is deliberate, and the 2023 Manual flattens it. AP I uses “considered” for persons and “presumed” only for a narrower class of objects. The 2023 Manual uses “presume” for both. The 1977 Diplomatic Conference replaced “presume” with “considered” in the persons rule precisely to avoid importing the evidential-burden connotations of “presumption” into that context (see Section I above; CDDH/50/Rev.1, p. 239, para. 39). The 2023 Manual undoes that choice for U.S. purposes.
Second, the political pressure preceding the revision was framed in “presumption” language. On 14 February 2023, Representative Sara Jacobs and Senator Dick Durbin wrote to the DoD General Counsel demanding a “clear statement of the legally required presumption of civilian status”. The term of art (presumption” was set before the redrafting started.
Third, the prior position was so extreme that a surgical revision was not available. The pre-2023 § 5.5.3.2 had denied the rule outright. To climb back to a defensible position, the DoD could not migrate to “considered for persons, presumed for objects”; it had to take the position the rules’ critics had spent a decade advocating (Hathaway, Lederman and Schmitt 2016; Goodman 2022), and that position was uniformly framed as a “presumption.”
Fourth, the 2023 revision broadens the doubt rules’ triggers. AP I confines the doubt rules to specific situations: Article 50 (1) AP I engages only “in case of doubt whether a person is a civilian”, and Article 52 (3) AP I engages only “in case of doubt” whether a specific class of object is being used militarily. The 2023 DoD Manual converts both rules into a single default starting point: every person or object is presumed protected from the outset, regardless of whether doubt has actually arisen. As Nasu and Watts observe, the 2023 formulation almost reads as “a universal ‘going in’ proposition” rather than a means to resolve doubt. The practical implication, on their reading, is that “[o]nly after disproving that initial presumption may a soldier carry out an attack.” Boothby separately identifies a related risk: that the 2023 wording collapses “what are, in law, two distinct rules” — the doubt rules and the underlying rules defining civilians and civilian objects. The stronger “presume” verb both signals and reinforces this broadening.
Read against AP I, the four causes converge on the same doctrinal feature: the 2023 DoD Manual asks more of the commander than the treaty requires.
C. What the test actually demands
The 2023 DoD Manual itself is careful about what the test does not require. The law of war does not impose a fixed standard of evidence or proof on the commander; what is reasonable depends on the circumstances of the attack (DoD Manual, at § 5.4.3.2). The relevant circumstances include the time and resources reasonably available, the risks to civilians from an erroneous decision, risks to friendly forces, and the military advantage expected from the attack (same provision).
Among the IHL experts who have written on the revision of the DoD Law of War Manual, the response is less unified than the strongest critique suggests. Geoffrey Corn treats the revision as “unremarkable” in operational substance but flags a structural concern: a presumption ordinarily takes its value from a defined rebuttal standard, and the 2023 Manual’s reliance on “good faith” leaves that gap unfilled. William Boothby supports the revision overall while observing that “it is the use of the word ‘presume’ that causes the potential ambiguity” between the doubt rules and the underlying rules defining civilians and civilian objects. Michael Schmitt frames the operative question without reaching for a fixed threshold: the issue is whether “the attacker is confident enough in the circumstances to mount the attack lawfully”.
Read together, the practitioner concerns track the doctrinal feature identified above without contesting the revision wholesale. The stronger verb imports rebuttable-presumption connotations into a context for which AP I deliberately chose less burdensome language. Corn’s missing-rebuttal-standard concern and Boothby’s drafting-ambiguity concern point at the same source. Schmitt’s framing — confidence in the circumstances rather than threshold-crossing — is the formulation that survives all the verbal divergence.
The reasonableness inquiry asks whether the commander reasonably executed the means available to verify the target, and whether the assessment was reasonable on the information he had. The means available shape what counts as reasonable execution: a commander who declines to consider relevant and reliable information at his disposal has not reasonably executed the verification process, and the ex post defensibility of his decision suffers accordingly. The full doctrinal analysis of particular means — staff input, intelligence products, decision-support tools — exceeds the scope of this post and belongs in dedicated treatment elsewhere in the series. What the doubt rules require is the commander’s reasonable engagement with what is available.
Section IV turns to ex post review — how the reasonable-commander standard reconstructs the ex ante judgement.
IV. The convergence with the reasonable-commander standard
The doubt rules do not live in isolation. They connect, at the moment of review, with the reasonable-commander standard that post #19 developed at length. The connection is doctrinal, not decorative: the doubt rules and the reasonable-commander standard operate together, with the former shaping what the commander must do under uncertainty and the latter shaping how a tribunal later reviews what he did.
A. Where the doubt rules meet the reasonable-commander standard
The temporal vantage point is the hinge. As Section II already showed, the Galić Trial Chamber tied the doubt rule for persons to the circumstances of the person contemplating the attack and the information then available to him (Galić Trial Judgment, at para. 50). That formulation is ex ante by construction. It asks what a person in the commander’s situation could reasonably have believed at the moment of decision, judged from the commander’s vantage point rather than reconstructed through later inquiry.
The reasonable-commander standard, developed in post #19, occupies the same vantage point from the reviewing tribunal’s side. The ICTY framed the inquiry around a reasonably well-informed person in the commander’s circumstances (Galić Trial Judgment, at para. 58), and located the relevant information at the time of the attack rather than at the time of judgement. Post #19 develops this standard in detail, including its limits as illustrated by the Gotovina Appeals Chamber’s review of margin-of-error inferences. I do not repeat that treatment here. What matters for present purposes is the structural fit.
B. The procedural form of ex post review
In my view, the doubt rules and the reasonable-commander standard meet at the same point. The doubt rules shape how the commander must reason when status doubt survives a reasonably executed inquiry. Article 50 (1) AP I directs that the person, in case of doubt, be considered a civilian; Article 52 (3) AP I directs that the object normally dedicated to civilian purposes be presumed not used for military purposes. Neither rule requires that residual doubt translate automatically into non-engagement. Both require that the assessment, on the information then available, be reasonable in the circumstances of the attack. The reasonable-commander standard, for its part, tells the tribunal how to review what the commander did — by asking whether a reasonably well-informed commander, given the means and information then available, could have reached the same decision. The inquiry anchors at reasonableness in the circumstances on both sides.
The practitioner consequence is concrete. Ex post review of a case engaging the doubt rules is procedural in form. The tribunal’s central question concerns the verification process and the reasonableness of the assessment on the information then available.
This convergence is the doctrinal close-out of the “deciding under uncertainty” thread that post #19 opened. The reasonable-commander standard answers how the commander reaches a sound assessment. The doubt rules answer what the law expects when his information is incomplete — without dictating that engagement is precluded simply because doubt remains. The two operate in series, and an ex post tribunal applies them in that order.
Conclusion
The doubt rules are narrower than the 2023 DoD Law of War Manual makes them. Article 50 (1) AP I uses “considered” for persons. Article 52 (3) AP I uses “presumed” for a narrower class of objects. Both rules engage only in case of doubt, after a reasonably executed verification. Neither requires the commander to eliminate all doubt before engaging a target. The 2023 Manual’s broader formulation — universal default, single verb, attached to the principle of distinction itself — asks more of the commander than the treaty does. The reasons are traceable to the political pressure preceding the revision and to the difficulty of climbing back from a position that previously denied the rules outright.
The corrective reading is procedural in form. The doubt rules engage when status doubt survives a reasonably executed inquiry. They require a reasonable assessment in the circumstances of the attack, on the information then available. The reasonable-commander standard, applied ex post, reviews that assessment by the same lights. The two operate in series. Together they form the legal framework for deciding under uncertainty.