Estimated reading time: 16 minutes
The duty of constant care is one of the most underrated obligations in the law of armed conflict. It sits in the first paragraph of Article 57 of Additional Protocol I to the Geneva Conventions of 1949 (AP I), beneath a heading that promises something narrower: “Precautions in attack.” Yet the paragraph reaches past attacks. It fixes the duty to “the conduct of military operations” as a whole and requires that constant care be taken to spare civilians and civilian objects. The gap between that heading and that text is where this post begins.
The post makes a claim some readers will find surprising. Constant care under Article 57(1) is not a decorative preamble to the specific precautions that follow it. It is a standing, operations-wide obligation with legal content of its own. The harder question is how much doctrinal weight it carries. Is it a rule tucked among the targeting rules, or an overarching principle that shapes the whole operation? That question, rule or principle, runs through everything below.
Article 57 is familiar ground for this series. I examined the specific precautions in attack in two earlier posts, the first on kinetic operations and the second on cyber operations. This post turns to the general obligation that sits above those specific precautions. It also follows a thread from my most recent post, which asked whether commanders must consider, verify and rely on the output of AI decision-support systems: constant care is the continuous duty within which such questions of machine-assisted judgement arise. I write as a practitioner, and with a leaning the reader should know. I am sceptical of the claim that modern technology has made civilian protection harder, and more interested in what the law actually asks of a commander who takes it seriously. The argument runs in five steps, from what the text says to where the duty’s limits lie.
I. A deceptively simple text
The duty of constant care is easy to underestimate. Article 57 of AP I is headed “Precautions in attack,” and most of what the Article contains is indeed about attacks. Its first paragraph is framed differently, and the difference is deliberate. Article 57(1) attaches the duty to “the conduct of military operations,” not to “the attack,” and requires that “constant care shall be taken to spare the civilian population, civilians and civilian objects.” Those opening words set the reach of the whole obligation.
That reach is wider than the heading suggests. Jensen reads the first subparagraph as one that “takes a much broader approach than just ‘attack’” (at p. 202). It binds armed forces to a general requirement even when they are not attacking (at p. 203). The ICRC Commentary supplies the reason: it treats “military operations” as any movements, manoeuvres and other activities that armed forces carry out with a view to combat (at para. 2191). Combat operations are a larger category than attacks, and Article 57(1) fastens the duty to the larger one.
The treaty text and its Commentary point the same way. The breadth of Article 57(1) is a property of the text, not an accident of its placement above the attack rules.
This is why the post holds two things apart from the outset. Constant care under Article 57(1) governs the conduct of military operations as a whole. The specific precautions in attack under Article 57(2) address the narrower situation of an attack. The two are closely related, and much turns on how that relationship is understood. In operational terms, the scope difference is not academic. Because the duty is “constant”, it runs in the approach to an attack, through the attack itself, and onward as operations continue. The moment of attack is therefore not the wrong time to exercise care. It is simply not the only time.
That leaves the harder question. Does the operations-wide duty carry independent legal weight, or is Article 57(1) only an introduction to the real rules that follow? The next section confronts it.
II. The chapeau thesis, fairly stated
Before defending the operations-wide reading, I want to state the case against it at full strength. Call it the chapeau thesis. On this account, Article 57(1) is an opening flourish: it announces a principle, and the real legal work is done by the specific precautions that follow in Article 57(2). I take the thesis seriously, and a careful reader should too, because respectable authority can be read to support it.
Quéguiner puts the perception plainly, even as he sets out to rebut it. Because the constant-care duty stands at the head of the Article in broad language, it is often seen as “merely inspirational, particularly in the light of its very general wording” (at p. 796). On this view, paragraph 1 acquires legal weight only in conjunction with one of the concrete rules beneath it.
The thesis draws surface support from the ICRC’s own materials. As Quéguiner notes, the Commentary presents the paragraphs after 57(1) as the practical application of the constant-care principle, which invites the inference that the enforceable content lives in the specifics. The customary-law study points the same way. It restates the duty as a single rule, Rule 15, that fuses constant care with the obligation to take all feasible precautions, and its commentary calls this a basic rule to which the following rules give content (Vol I, at p. 51).
The thesis is strongest in its practical form. When a tribunal or a targeting review asks whether a party took the precautions the law requires, it reaches for the specific duties: verify the objective, choose means and methods that spare civilians, cancel or suspend an attack that turns unlawful, and warn where circumstances permit. Those duties are concrete enough to apply and to enforce. A general instruction to take constant care can seem, by comparison, too open-textured to decide a case on its own. On the chapeau thesis, that is precisely why paragraph 1 sits where it does: it orients the specifics without adding an obligation of its own.
Stated this way, the thesis is no straw man. It is a coherent account of how Article 57 is built and how precautions are litigated. If the operations-wide reading is to hold, it has to answer the thesis on that ground. The next section takes up the challenge.
III. Why the duty of constant care is a standing, operations-wide obligation
The reading the sources converge on
The chapeau thesis fails on its own ground. The authorities it leans on do not treat Article 57(1) as an empty preface. They treat it as a standing obligation with content of its own. Quéguiner sets out the perception that paragraph 1 is a mere inspirational preamble, only to reject it. The family resemblance between paragraph 1 and the specific rules that follow does not make the two identical, and it does not strip paragraph 1 of independent legal effect (at p. 796).
The Commentary the thesis cites says as much. It describes Article 57(1) as “a general principle which imposes an important duty on belligerents” (at para 2191). The Commentary does also call the paragraph a restatement of a principle that customary law already contains. That speaks to the duty’s pedigree, not its force. A principle that imposes an important duty is no decorative heading.
That customary pedigree sets the reach of the obligation. The ICRC’s study on customary international humanitarian law carries constant care as a rule in its own right, Rule 15, binding in both international and non-international armed conflict. The duty therefore does not wait on treaty membership. It binds as custom.
A serving state’s military lawyer reaches the same reading. Writing in a personal capacity, Neuman reads the constant-care duty as reaching all military operations, not only attacks, despite the attack-focused title of Article 57 (at p. 77). That is the operations-wide reading, stated from inside the practitioner community the thesis is meant to reassure.
The convergence across these treatments is what defeats the thesis. The Commentary and the customary-law study, the very authorities the narrow reading invokes, describe paragraph 1 as a duty-imposing principle with customary standing. Far from supporting the thesis, they cut against it. Article 57(1) is a standing, operations-wide obligation. Whether states and their militaries actually treat it that way is the next question.
What states and militaries actually do
The operations-wide reading is not a scholars’ preserve. States and their armed forces treat constant care as a standing obligation that reaches beyond the moment of attack. Consider the United States, which is not even a party to AP I. Its Law of War Manual requires feasible precautions in the conduct of attacks (at para 5.11). It does not stop there. The Manual recognises precautionary duties outside the conduct of attacks, in defence planning and other military operations, to limit harm to civilians and other protected persons (at para 5.14). A non-party that writes these obligations into its own doctrine, and that recognises precautionary duties beyond the attack, is not reading Article 57(1) as an empty preface.
Germany, an AP I party, points the same way. Its current Handbook on International Humanitarian Law (Zentrale Dienstvorschrift A-2141/1) carries the constant-care duty into its section on precautionary measures for air operations. It provides that in air warfare operations too, “ist ständig darauf zu achten” (care must be taken at all times) to protect civilians and civilian objects from the effects of hostilities (at para. 1153, p. 145). The manual does not use the treaty’s English wording, yet that phrase is the German rendering of constant care, and the word “too” marks the duty as general: it holds in the air as it holds in other operations.
NATO doctrine goes further still. Its joint-targeting doctrine, AJP-3.9(B), lists precaution among the principles that govern targeting and treats it as running through the whole process rather than arising at the point of engagement (Section 1.2.3, at p. 1-4). Legal advisers are embedded across the cycle, from the point a target is first identified through its validation and eventual engagement (Section 1.6.1, at p. 1-23). Precaution, in NATO’s account, is a discipline that spans the operation.
The pattern is consistent across these systems. A treaty party and a non-party, a national manual and an alliance doctrine, describe the same continuous obligation. What they debate is how far constant care reaches and how it is discharged, not whether it binds before and beyond the attack.
Rule or principle, and why the label matters
The real dispute is not whether constant care binds beyond attacks. It does. The dispute is how much doctrinal weight it carries: a rule tucked among the targeting rules, or an overarching principle that shapes the whole operation. The demanding side has serious champions. Writing on the US Law of War Manual, Geoffrey Corn argues that the precautions obligation is broader than a mere rule and belongs among the fundamental principles of the law of armed conflict. Constant care, on his account, “must animate all strategic, operational, and tactical decision-making,” and treating Article 57 as merely a targeting-section rule sells the obligation short. He reads the principle framing as the more accurate description of how the best commanders work, mitigating civilian risk well before the proportionality question arises.
Corn is not alone in reaching for the language of principle. As already noted, the ICRC Commentary on Article 57 casts Article 57(1) as a general principle (at para 2191), and NATO doctrine lists precaution among its targeting principles. The vocabulary of principle is not an activist import; it sits in the ICRC Commentary and in alliance doctrine.
In my view, the label matters less than practitioners sometimes fear. Whether constant care is filed as a rule or elevated to a principle, its operational content is the same continuous discipline: keep civilian protection in view across the operation, not only at the trigger. The principle framing is simply the more honest description of practice, and it guards against the misreading I flagged at the outset of this post, treating constant care as a box ticked at the moment of attack. That is the payoff of the operations-wide reading. Whether it makes concrete demands a commander can act on is the question the next section answers.
IV. What the duty of constant care demands in practice
An operations-wide duty is only as good as its content. If constant care meant no more than a pious wish to spare civilians, the chapeau thesis would be half right. It means more than that. The duty resolves into identifiable tasks that run through an operation, and military doctrine already names them.
The first demand is situational awareness that is maintained, not sampled once. Jensen reads constant care as barring a commander from ignoring the effects of operations on the civilian population, and as requiring that awareness to be kept current as an operation unfolds (at pp. 202–204). His worked examples come from cyber operations, but the reading of Article 57(1) is domain-general: the duty to keep civilian effects in view does not switch off between the plan and the act.
Doctrine locates this work in a specific place. In NATO’s joint targeting cycle, the targeting staff (the multidisciplinary element of intelligence, operations, and legal advisers that develops, validates, and recommends targets to the commander) carries the duty in practice (AJP-3.9(B), Section 1.5.2, at p. 1-21). Legal advisers are not summoned at the end to bless a finished plan. They act from the beginning, so that IHL/LOAC principles are integrated across the whole process, from the first identification of a target through its validation and eventual engagement (Sections 1.6.1 and 1.6.2, at p. 1-23 et seq.).
The duty has teeth at the moment of action. NATO doctrine requires commanders to take every feasible step to confirm that a target is a lawful military objective before engaging it (Section 1.7, at p. 1-25). Those carrying out an engagement must act on the facts available to them, and on the facts they should reasonably have obtained, and must cancel or suspend the engagement if it emerges that the target is not a military objective or that the expected collateral damage would be excessive, in the circumstances ruling at the time (Section 1.7, at p. 1-25). Constant care is what makes cancellation thinkable: a commander who has kept civilian protection in view will recognise the reason to stop.
It also reaches past the strike. NATO’s assessment phase builds in a civilian-casualty tracking mechanism, so that what happened in one engagement informs future operations and further reduces civilian harm (Section 1.5.1, f. at p. 1-19). This is not a free-standing post-strike rule. It is the operation’s habit of learning from its own effects, which is how the continuity of the duty shows up once a given engagement is over.
Decision-support tools can help carry this load. Collateral-damage estimation, and the AI decision-support systems I examined in an earlier post, can sharpen a commander’s picture of expected harm. NATO is careful about their status: the methodology is meant to “aid and not replace the judgement of the commander” (Section 1.7, at p. 1-28). The tool informs the duty; it does not discharge it.
The practitioner reading is that constant care is a working discipline, not an aspiration. It asks for continuous awareness, a staff organised to act on it, the readiness to stop, and the habit of learning afterward. None of that is beyond a competent headquarters. Whether the law asks too much when the information is poor or the tempo is high is the question the final section takes up.
V. The boundary: feasibility and defence readiness
Constant care is demanding, but it is not boundless. Two limits keep it a workable standard rather than a counsel of perfection. The first is feasibility, which fixes the ceiling on what the duty can require. The second is that legitimate military considerations, defence readiness among them, count within the feasibility judgment itself.
Feasibility is the fulcrum. On ratifying AP I, the United Kingdom recorded its understanding that “feasible” means “practicable or practically possible, taking into account all circumstances ruling at the time” (lit. (b)), adding that those circumstances include humanitarian and military considerations. NATO’s targeting doctrine adopts the same definition and observes that some states speak of “feasible precautions” rather than “all feasible precautions” (AJP-3.9(B), Section 1.5.1, at p. 1-18, fn. 28). The duty is measured by what is practicable in the circumstances. Perfect protection is not the test. Absolute certainty about a target is never required, and there is no duty to pursue endless verification (Neuman, at p. 78).
The measure is applied looking forward, not backward. The United Kingdom’s declaration also records that commanders decide on the information reasonably available to them at the relevant time (lit. (c)). Neuman makes the same point from the bench’s perspective: cases are not to be judged in hindsight (at p. 81). Compliance turns on the quality of the decision on the information at hand, not on how the attack happened to turn out. A good-faith, diligent decision that proves imperfect is not, for that reason, a breach.
Defence readiness sits inside this judgment, not outside it. Constant care does not require a commander to give civilians precedence over reasonable military requirements; it requires sensitivity to civilian effects and their mitigation where feasible (Neuman, at p. 77). Neuman states the operational point directly: “Force preservation is a legitimate military consideration” (at p. 81). Where sending ground forces in would expose them to markedly higher risk, a commander may reasonably conclude that this option is not feasible. The point has judicial backing. Reviewing the NATO air campaign, the ICTY committee found nothing inherently unlawful in flying above the reach of enemy air defences, so long as commanders still do what is practicable to distinguish military objectives, which it judged had been achieved in the great majority of cases despite the 15,000-foot floor (Final Report to the Prosecutor by the Committee Established to Review the NATO Bombing Campaign, at para. 56). United States doctrine makes a comparable allowance. In its 2023 Law of War Manual update, the Department of Defense confirmed that commanders may act “at the speed of relevance,” even in high-intensity conflict, so long as they proceed in good faith on the information they have at the time (Section 5.4.3.2, at p. 204).
The practitioner reading is that constant care asks for diligent, good-faith judgement. It does not demand omniscience, and it does not require a commander to endanger his own force. It is a continuous discipline, bounded by what is practicable and by the legitimate demands of the mission and the safety of one’s own troops. That is what makes the operations-wide reading defended in this post realistic rather than utopian: it asks commanders to keep civilian protection continuously in view and to act on it so far as they reasonably can. That is a great deal, and it is also the limit.
Conclusion
The duty of constant care deserves better than its reputation as a preamble. Article 57(1) attaches a real and continuous obligation to the conduct of military operations, and the specific precautions in Article 57(2) are one expression of it, not the whole of it. The chapeau thesis, which reads the first paragraph as throat-clearing before the operative rules, understates both the treaty and the practice built on it.
Whether we classify constant care as a rule or elevate it to a principle matters less than what it asks of a commander, which is the same either way. Keep civilian protection continuously in view, across the operation and through each engagement, rather than only at the moment of attack.
That duty is demanding without being unrealistic. It is measured by what is feasible in the circumstances ruling at the time, assessed on the information reasonably available rather than with hindsight, and it leaves room for legitimate military considerations, among them the safety of one’s own force and the tempo an operation demands. Constant care asks a commander to try, continuously and in good faith, to spare civilians so far as is practicable.
If constant care is the continuous discipline described here, someone has to make it work day to day. Increasingly that falls to the legal adviser, whose availability Article 82 AP I requires and whose task grows more complex as planning and targeting draw on AI decision-support systems. My next post takes up the legal adviser under Article 82 in AI-enabled operations.
For now, the point is a simple one. Constant care is Article 57’s standing demand, alive from the planning of an operation through its execution and after. The heading understates it. A commander should not.