How weapons law emerged

Dr Francis Lieber ⁵ prepared a Code issued to the Union side in the American Civil War in which he explained military necessity as the necessity of those measures which are indispensable for securing the ends of the war. He considered that military necessity should not permit cruelty, such as the infliction of suffering for its own sake, or the use of poison. ⁶ Although the Lieber Code was never adopted by states and is therefore not regarded as a source of international law as such, it does represent an informed assessment of law as it then existed. The prohibition of the use of poison is also reflected in Article 23(a), Hague Regulations 1907, ⁷ and is certainly a rule that applies to all States and in all armed conflicts, international and non-international.

The idea that the infliction of suffering for its own sake is prohibited later developed into one of the two core principles of the law of weaponry which prohibits weapons that by nature cause superfluous injury or unnecessary suffering. We will discuss this principle in a later section of this chapter.

A Declaration adopted in St Petersburg in 1868 prohibited ‘the employment … of any projectile of a weight below 400 grammes, which is either explosive or charged with fulminating or inflammable substances’. ⁸ It is fair to say that the weight limit prescribed by the treaty no longer reflects the law.

Customary law does, however, continue to prohibit the use of explosive or incendiary bullets designed exclusively for use against personnel. This is because a solid round would achieve the relevant military purpose, so using an explosive round against such targets will cause injury for which there is no military necessity. ⁹

The Peace Conferences conducted in The Hague in 1899 and 1907 took the development of weapons law further. Declaration II prohibited the use of ‘projectiles the sole object of which is the diffusion of asphyxiating or deleterious gases’, ¹⁰ but only applied to the states party in the case of a war between two or more of them and would not apply if a state that was not a party took part in the conflict, and the Declaration did not prohibit weapons which combined gas diffusion with some other object, such as blast or fragmentation. The Declaration did not therefore prevent the gas attacks on the trenches during the First World War that resulted in so many casualties and so much suffering. It was in the aftermath of this experience that the Geneva Protocol of 1925 was adopted. Many decades later, the Chemical Weapons Convention of 1993 and the Biological Weapons Convention of 1972 addressed these mass casualty weapons in an altogether more comprehensive way.

The third Declaration of 1899 prohibited ‘the use of bullets which expand or flatten easily in the human body, such as bullets with a hard envelope which does not entirely cover the core or is pierced with incisions’. ¹¹ It also only applied in the case of a war between two or more states that are party to the treaty and would not apply if a non-contracting state entered the conflict. The prohibition now has customary law status, meaning that it binds all states whether or not they are party to the treaty. The customary rule prohibits the use in armed conflicts between states of bullets that are designed, in the intended circumstances of use, to expand or flatten easily in the human body.

The position in relation to armed conflicts within a state is more complex. Consider, for example, a hostage situation in which a hostage-taker must be instantly disabled for the protection of civilians, or consider a situation in which the ricochet to be expected from a high velocity round will imperil civilians in the vicinity of the target. The use of expanding ammunition in such relatively limited non-international armed conflict situations would cause the target additional suffering for which there is a military purpose, namely the protection of the relevant civilians, so the superfluous injury/unnecessary suffering rule would not, arguably, be breached in those limited kinds of circumstance and, therefore, neither would international law. ¹²

Superfluous injury and unnecessary suffering

At this point we should conclude our brief account of the early emergence of the law of weaponry and start to get to grips with its two core principles. The first of these prohibits injury that lacks military utility, an idea first expressed in the modern era in the Preamble to the St Petersburg Declaration 1868:

Considering that the progress of civilisation should have the effect of alleviating as much as possible the calamities of war;

That the only legitimate object which states should endeavour to accomplish during war is to weaken the military forces of the enemy;

That for this purpose it is sufficient to disable the greatest possible number of men;

That this object would be exceeded by the employment of arms which uselessly aggravate the sufferings of disabled men, or render their death inevitable;

That the employment of such arms would therefore be contrary to the laws of humanity. ¹³

The Brussels Declaration and the Oxford Manual contain similar statements, ¹⁴ but it was in Article 23(e) of the Hague Regulations of 1907 that the notion was first expressed as a substantive rule of law, as follows: ‘It is especially forbidden to employ arms, projectiles or material calculated to cause unnecessary suffering.’

Seventy years later, in API, the modern formulation of the rule provides:

It is prohibited to employ weapons, projectiles and material and methods of warfare of a nature to cause superfluous injury or unnecessary suffering. ¹⁵

This customary rule, which binds all states in all types of armed conflict, is central in importance to the law of weaponry. The practical application of the rule, however, involves a comparison of inherently dissimilar phenomena. The elements to be compared are, it is suggested, the degree and relative extent of the suffering or injury that the use of the weapon will inevitably occasion and the generic military advantage or utility to be anticipated from the employment of the given weapon in its intended circumstances of use. The terms ‘superfluous’ and ‘unnecessary’ confirm the comparative nature of the test that is to be applied, leading to the conclusion that a weapon is likely to breach the rule if it may be expected to cause injury on a scale significantly greater than that to be expected of alternative weapons that yield the same generic military advantage or utility.

Such a comparison process only makes sense if it is the employment of the weapon in its normal, designed circumstances of use that is evaluated. ¹⁶ Perfectly lawful weapons are capable of being misused and of having unacceptable or even unlawful effects as a result of such misuse. The proper basis for judgement is therefore how the weapon behaves when used in its normal circumstances, within its designed range, employing its intended power setting or velocity, and when being directed at its intended category of target.

So the principle can be expressed as follows:

the legitimacy of a weapon, by reference to the superfluous injury and unnecessary suffering principle, must be determined by comparing the nature and scale of the generic military advantage to be anticipated from the weapon in the application for which it is designed to be used, with the pattern of injury and suffering associated with the normal intended use of the weapon. ¹⁷

It is for states to evaluate which weapons breach the rule and to assess new weapons and weapon technologies accordingly.

Indiscriminate weapons rule

The second fundamental customary principle of the law of weaponry prohibits weapons whose nature it is to be indiscriminate. During the period before 1974 a rule that indiscriminate attacks are prohibited certainly existed, but there was no treaty text specifically addressing weapons that by nature are indiscriminate. An authoritative commentator noted in 1975 that when the negotiations of what was to become API commenced, not all experts were prepared to acknowledge that a rule prohibiting identifiable ‘indiscriminate weapons’ had ‘acquired the status of a rule of positive international law’. ¹⁸ API prohibited indiscriminate attacks ¹⁹ which it defined as including attacks ‘which employ a method or means of combat which cannot be directed at a specific military objective; or … which employ a method or means of combat the effects of which cannot be limited’ as required by the Protocol and which consequently are of a nature to strike military objectives and civilians or civilian objects without distinction. The clear references in the rule to means of combat render the pre-existing indiscriminate attacks rule into a rule that applies specifically and explicitly to weapons.

The V2 rocket used to attack Southern England from September 1944, ²⁰ its predecessor the V1 rocket, and some Scud rockets have been cited as examples of weapons that would have breached the rule.

Geneva Gas Protocol 1925

The perceived inadequacies of the Hague Declaration IV of 1899 were addressed in 1925 with the adoption of the Geneva Gas Protocol. ²¹ This treaty prohibited the use in war of asphyxiating, poisonous, or other gases and of all analogous liquids, materials, or devices, and it extended the prohibition to the use of bacteriological methods of warfare. The ‘sole object’ language, which had provided the loophole in the 1899 text, was not repeated in the 1925 Protocol.

France, the UK and US became party to the Protocol on the stated understanding that they would only remain bound by the prohibitions so long as the adverse party in an armed conflict did not use the prohibited weapons. These ‘no first use’ arrangements remained relevant until the adoption of the Chemical and Biological Weapons Conventions. ²²

Weapons and the environment

The next relevant piece of conventional law came some 50 years later with the adoption of the Environmental Modification Convention of 1976 (ENMOD). ²³ This treaty prohibited military or other hostile use of environmental modification techniques if these would have widespread, long-lasting or severe effects as the primary means of destruction, damage or injury to another state party. ²⁴ This provision will therefore need to be considered if a weapon uses the environment as an instrument for causing the stated degree of damage to another state that is party to the treaty.

The other environmental protection rule within the law of armed conflict is to be found in Articles 35(3) and 55 of API. Article 35(3) prohibits the employment of methods or means of warfare which are intended, or may be expected, to cause widespread, long-term, and severe damage to the natural environment. Article 55 builds on this rule by imposing a requirement to take environmental care and by expressly prohibiting methods or means that are intended or may be expected to cause the prohibited damage and thereby to prejudice the health or survival of the population. Such a requirement to take care will clearly apply to the selection and design of weapons. The API rules are, however, concerned with collateral damage to the environment, as opposed to its use as a weapon.

Conventional Weapons Convention of 1980

Further progress in developing the law of weaponry was achieved with the adoption of the Convention on Prohibitions or Restrictions on the Use of Certain Conventional Weapons which May be Deemed to be Excessively Injurious or to have Indiscriminate Effects, which we shall refer to as the CCW. ²⁵ This Convention facilitates the negotiation and adoption of Protocols that address particular types of weapon. Some of the provisions in these Protocols merely restate the customary principles that we have already discussed or apply to the particular kind of weapon the established targeting rules that apply as a matter of law to all attacks. The following summary will focus on the provisions that seem to develop the law of weaponry.

CCW Protocols prohibit the following specific weapons:

Protocol 1 prohibits the use of ‘any weapon the primary effect of which is to injure by fragments which in the human body escape detection by x-rays’. ²⁶

Protocol II ²⁷ and Amended Protocol II ²⁸ address the use in armed conflict of mines, ²⁹ boobytraps, and other devices. ³⁰ Protocol II defines booby-traps as ‘any device or material which is designed, constructed or adapted to kill or injure and which functions unexpectedly when a person disturbs or approaches an apparently harmless object or performs an apparently safe act’. ³¹ Under Article 6 of Protocol II it is prohibited to use booby-traps in the form of an apparently harmless portable object if they are specifically designed and constructed to contain explosive material and to detonate when they are disturbed or approached. ³² The use of booby-traps is also prohibited if they are in any way attached to or associated with:

• internationally recognised protective emblems, signs or signals;
• sick, wounded or dead persons;
• burial or cremation sites or graves;
• medical facilities, medical equipment, medical supplies or medical transportation;
• children’s toys or other portable objects or products specially designed for the feeding, health, hygiene, clothing or education of children;
• food or drink;
• kitchen utensils or appliances except in military establishments, military locations or military supply depots;
• objects clearly of a religious nature;
• historic monuments, works of art or places of worship which constitute the cultural or spiritual heritage of peoples;
• animals or their carcasses. ³³

Amended Protocol II prohibits:

• mines, booby-traps or other devices which employ a mechanism or device specifically designed to detonate the munition by the presence of commonly available mine detectors as a result of their magnetic or other non-contact influence during normal use in detection operations; ³⁴
• a self-deactivating mine equipped with an anti-handling device that is designed in such a manner that the anti-handling device is capable of functioning after the mine has ceased to be capable of functioning; ³⁵
• anti-personnel mines that do not incorporate in their construction a material or device that enables the mine to be detected by commonly available technical mine detection equipment and provides a response signal equivalent to a signal from 8 grammes or more of iron in a single coherent mass; ³⁶
• remotely delivered anti-personnel mines which do not comply with the following requirements: ³⁷ their design and construction must be such that no more than ten percent of activated mines will fail to self-destruct within 30 days after emplacement, and each mine must have a back-up self-deactivation feature designed and constructed so that, in combination with the self-destruction mechanism, no more than one in 1000 activated mines will function as a mine 120 days after emplacement; ³⁸
• remotely delivered mines other than anti-personnel mines, unless, to the extent feasible, they are equipped with an effective self-destruction or self-neutralisation mechanism and have a back-up self-deactivation feature, which is designed so that the mine will no longer function as a mine when the mine no longer serves the military purpose for which it was placed in position; ³⁹
• and booby-traps or other devices in the form of apparently harmless portable objects which are specifically designed and constructed to contain explosive material. ⁴⁰

Protocol IV to CCW, which owes much to the general disapproval of blinding as a method of warfare, provides:

It is prohibited to employ laser-weapons specifically designed, as their sole combat function or as one of their combat functions, to cause permanent blindness to unenhanced vision, that is, to the naked eye or to the eye with corrective eyesight devices. ⁴¹

At the time of writing, it can be concluded that a customary rule in the terms of Protocol I to CCW is in the process of emerging and that a customary rule in the precise terms of article 1 of Protocol IV has emerged.

Bacteriological or biological weapons

The 1925 Geneva Gas Protocol only prohibited the use of bacteriological or biological weapons. Article I of the Biological Weapons Convention ⁴² takes matters considerably further, in that the states party undertakes never in any circumstances to develop, produce, stockpile, or otherwise acquire or retain:

1. Microbial or other biological agents or toxins whatever their origin or method of production, of types and in quantities that have no justification for prophylactic, protective or other peaceful purposes;
2. Weapons, equipment or means of delivery designed to use such agents or toxins for hostile purposes or in armed conflict.

The Convention does not explicitly prohibit ‘use’, but at the Fourth Review Conference in 1996 it was agreed among the States party that Article 1 has the effect of prohibiting the use of such weapons. ⁴³

The comprehensive terms of the prohibitions mean that they apply to all classes of conflict. Very many states are party to the treaty, including almost all militarily significant states, and the consistent practice of states supports the conclusion that the prohibition on use of such weapons is now a customary rule of international law which therefore binds all states, whether or not they are party to this treaty. ⁴⁴ It is also probably the case that the treaty’s prohibition on the possession, stockpiling, transfer and development of such weapons is also customary in nature. So it is unlawful for any state, whether it is party to the Biological Weapons Convention or not, to plan, prepare for, equip itself for, or undertake a biological attack.

While attempts have been made at two Review Conferences to agree upon a verification mechanism for the Convention, it has not, to date, proved possible to adopt such a provision.

Chemical weapons

The possession of chemical weapons was also not prohibited by the 1925 Gas Protocol. The adoption in 1993 of the Chemical Weapons Convention, ⁴⁵ which applies to all classes of conflict, therefore takes the law forward by providing that the participating states agree never under any circumstances:

• To develop, produce, otherwise acquire, stockpile or retain chemical weapons or transfer, directly or indirectly, chemical weapons to anyone;
• To use chemical weapons;
• To engage in any military preparations to use chemical weapons;
• To engage in any military operations to use chemical weapons;
• To assist, encourage or induce, in any way, anyone to engage in any activity prohibited to a State party under the Convention. ⁴⁶

‘Chemical weapons’ means, together or separately:

• Toxic chemicals and their precursors, except where intended for purposes not prohibited under this Convention, as long as the types and quantities are consistent with such purposes;
• Munitions and devices, specifically designed to cause death or other harm through the toxic properties of those toxic chemicals … which would be released as a result of the employment of such munitions and devices;
• Any equipment specifically designed for use directly in connection with the employment of munitions and devices specified in sub-paragraph (b). ⁴⁷

A toxic chemical is any chemical which, through its chemical action on life processes, can cause death, temporary incapacitation or permanent harm to humans or animals. This includes all such chemicals, regardless of their origin or of their method of production, and regardless of whether they are produced in facilities, in munitions, or elsewhere. ⁴⁸

A ‘precursor’ is any chemical reactant which takes part, at any stage, in the production, by whatever method, of a toxic chemical, including any key component of a binary or multi-component chemical system. ⁴⁹

If the chemical is intended for purposes which are not prohibited under the Convention and if the amount held is consistent with such innocent purposes, its possession is lawful. Purposes not prohibited under the Convention are:

• Industrial, agricultural, research, medical, pharmaceutical or other peaceful purposes;
• Protective purposes, namely those purposes directly related to protection against toxic chemicals and to protection against chemical weapons;
• Military purposes not connected with the use of chemical weapons and not dependent on the use of the toxic properties of chemicals as a method of warfare;
• Law enforcement, including domestic riot control purposes. ⁵⁰

The net effect of these provisions is, quite simply, to prohibit chemical warfare as that term is colloquially understood. The treaty applies in all classes of conflict and the prohibition on use of chemical weapons is clearly now a rule of customary law, ⁵¹ with the result that all states, irrespective of their participation in the Convention, are prohibited from using such weapons.

Riot control agents may not be used as a method of warfare, but they may be used for law enforcement, including domestic riot-control purposes. Riot control agents are ‘chemicals not listed in a Schedule to the Treaty which can produce rapidly in humans sensory irritation or disabling physical effects which disappear within a short time following termination of exposure’. ⁵² In the context of modern armed conflict, it is not of course always clear where law enforcement ends and armed conflict begins.

Anti-personnel landmines

Consensus is required before a CCW Protocol can be adopted, ⁵³ and no consensus could be achieved in favour of an Anti-Personnel Landmine (APL) ban. It was therefore an ad hoc process that led to the adoption of the Ottawa Convention, ⁵⁴ which provides:

Each state party undertakes never under any circumstances:

• To use anti-personnel mines;
• To develop, produce, otherwise acquire, stockpile, retain or transfer to anyone, directly or indirectly, anti-personnel mines;
• To assist, encourage or induce, in any way, anyone to engage in any activity prohibited to a State party under this Convention.

An anti-personnel mine is a mine designed to be exploded by the presence, proximity, or contact of a person and that will incapacitate, injure or kill one or more persons. Mines designed to be detonated by the presence, proximity, or contact of a vehicle as opposed to a person and that are equipped with anti-handling devices are not considered anti-personnel mines as a result of being so equipped. ⁵⁵

The treaty prohibitions apply in all classes of conflict, but have not yet achieved customary law status; they are therefore binding only on states party to the treaty.

Cluster munitions

The Convention on Cluster Munitions, adopted at a meeting in Dublin in 2008, obliges states party never under any circumstances:

• to use cluster munitions,
• to develop, produce, otherwise acquire, stockpile, retain or transfer to anyone, directly or indirectly, cluster munitions,
• to assist, encourage or induce anyone to engage in any activity prohibited to a State party under the Convention. ⁵⁶

A cluster munition means a conventional munition that is designed to disperse or release explosive sub-munitions each weighing less than 20 kilograms and includes those explosive submunitions. It does not mean the following:

• A munition or sub-munition designed to dispense flares, smoke, pyrotechnics or chaff, or a munition designed exclusively for an air defence role;
• A munition or sub-munition designed to produce electrical or electronic effects;
• A munition that, in order to avoid indiscriminate area effects and the risks posed by unexploded sub-munitions, has all of the following characteristics:
  • Each munition contains fewer than 10 explosive sub-munitions;
  • Each explosive sub-munition weighs more than 4 kilograms;
  • Each explosive sub-munition is designed to detect and engage a single target object;
  • Each explosive sub-munition is equipped with an electronic self-destruction mechanism;
  • Each explosive sub-munition is equipped with an electronic self-deactivating feature. ⁵⁷

‘Explosive sub-munitions’, ⁵⁸ ‘self-destruction mechanism’, and ‘self-deactivating mechanism’ are defined terms. Article 21 addresses the interoperability issues that may arise for states party engaging in military cooperation and operations with states not party to the Convention.

The rules in the Convention apply to states party only, but in all classes of conflict.

Non-international armed conflict

The scope of application of the Conventional Weapons Convention and its annexed Protocols was extended in 2001 ⁵⁹ to ‘situations referred to in Article 3 common to the Geneva Conventions of 12 August 1949’. ⁶⁰ The Protocols have that extended scope in relation to states that ratify that extension of 2001. APII always did apply to such conflicts. ⁶¹

The Biological Weapons, Chemical Weapons, Ottawa, and Cluster Munitions Conventions apply to all classes of conflict, as do the customary law on superfluous injury/unnecessary suffering and indiscriminate weapons principles and the customary rules of weapons law, such as the prohibition of poisons or poisoned weapons. The position in relation to expanding bullets was discussed earlier.

Treaty law restrictions on the use of certain weapons

Protocol II and Amended Protocol II to the CCW include restrictions on the use of mines, booby-traps and other devices the details of which lie outside the scope of this chapter. Protocol IV ⁶² also placed restrictions on the use of certain weapons that are not specifically prohibited.

Protocol III ⁶³ to the CCW defines an incendiary weapon as

any weapon or munition which is primarily designed to set fire to objects or to cause burn injury to persons through the action of flame, heat, or a combination thereof, produced by a chemical reaction of a substance delivered on the target. ⁶⁴ [It is prohibited] in all circumstances to make any military objective located within a concentration of civilians ⁶⁵ the object of attack by air-delivered incendiary weapons. ⁶⁶

It is also prohibited to make military objectives located within a concentration of civilians the object of attack using incendiary weapons other than air-delivered incendiary weapons

except when such military objective is clearly separated from the concentration of civilians and all feasible precautions are taken with a view to limiting the incendiary effects to the military objective and to avoiding, and in any event to minimizing, incidental loss of civilian life, injury to civilians and damage to civilian objects. ⁶⁷

The Protocol also prohibits making forests or other kinds of plant cover the object of attack using incendiaries except when such natural elements are used to cover, conceal, or camouflage combatants or other military objectives or have themselves become military objectives. ⁶⁸

Non-lethal weapons

NATO policy refers to non-lethal weapons as ‘weapons which are explicitly designed and developed to incapacitate or repel personnel, with a low probability of fatality or permanent injury, or to disable equipment with minimal undesired damage or impact on the environment’. ⁶⁹

The important point to note is that the ‘non-lethal’ character of a weapon does not affect the applicability to it of the principles and rules discussed so far in this chapter.

Weapon reviews

States that are party to API ⁷⁰ are required

[i]n the study, development, acquisition or adoption of a new weapon, means or method of warfare […] to determine whether its employment would, in some or all circumstances, be prohibited by th[e] Protocol or by any other rule of international law applicable to the High Contracting Party. ⁷¹

If, therefore, a weapon, means or method is being studied, then the weapons review duty applies. It will be a matter for national judgement when general technology research becomes the study of a weapon. ⁷²

‘Development’ involves the application of materials, equipment, and other elements to form a weapon and includes the improvement, refinement, and probably the testing of the prototype weapons with a view to achieving optimal performance. ⁷³ ‘Acquisition’ involves obtaining weapons from commercial undertakings and/or from other states and ‘adoption’ involves a state or its armed forces deciding to use a weapon or method of warfare in military operations.

Customary law requires all states to review new weapons to determine whether they comply with the law that applies to the relevant state. The final paragraph of the 1868 St Petersburg Declaration, Article 1 of Hague Convention IV of 1907, the AMW Manual, ⁷⁴ and the Tallinn Manual all lead to this conclusion. ⁷⁵ The ICRC concludes that ‘[t]he requirement that the legality of all new weapons, means and methods of warfare be systematically assessed is arguably one that applies to all States, regardless of whether or not they are party to Additional Protocol 1. ⁷⁶ However, relatively few states are known systematically to ensure the review of all new weapons. ⁷⁷

The law is not prescriptive as to the form that a weapon review must take, does not lay down any required procedure, and does not oblige states to disclose the contents of their reviews. Advice to an appropriate commander may be sufficient. ⁷⁸

The legal review must apply the rules of weapons law that apply to the state in question. These will be the two customary principles discussed in sections 3 and 4, the customary rules that have been identified, and the treaty prohibitions and restrictions to which the relevant state is party. The important point to note is that it is the principles and rules of current law that must be applied to new weapons and weapon technologies. The peculiarities of a new weapon technology are, therefore, no basis for arguing that an established weapons law rule does not have to be complied with.

Autonomous weapons

We have discussed the principles and rules of the law of weaponry and the obligation legally placed on all States to review new weapons. This section considers how these principles and rules apply to an emerging novel technology, namely autonomous weapons.

A weapon systems is ‘man on the loop’ if it is capable of automated or autonomous operation but is supervised by an operator who has the capability to intervene and override an attack decision that the automated or autonomous decision-making process makes. Contrast that to ‘man in the loop’ weapon systems in which the human operator decides which target is to be engaged and undertakes the attack by initiating the firing mechanism using the remote-control facility built into the weapon system. ⁷⁹

The weapon reviewer is concerned with the law applicable to the weapon system as such and not with the legality of a particular attack. He will therefore want to know whether a ‘man on the loop’ weapon system is capable of use in compliance with targeting law, including the distinction, discrimination, and proportionality principles and the precautions rules. If an individual empowered to countermand unsatisfactory machine-made attack decisions is enabled properly to supervise the autonomous or automated attack decision-making and to intervene when it is appropriate to do so, the autonomous or automated nature of the initial decision-making facility is unlikely to raise international weapons law concerns; ⁸⁰ i.e., such a weapon system is capable of being employed in compliance with the targeting rules.

It seems clear that the ultimate goal of much contemporary research is complete autonomous decision-making in attack. ⁸¹ There is, however, no internationally agreed legal definition of what automated and, respectively, autonomous attack decision-making means. ⁸²

Current UK doctrine refers to highly automated systems that ‘are constrained by algorithms that determine their responses by imposing rules of engagement and setting mission parameters which limit their ability to act independently’. ⁸³ Such a system is not remotely controlled, but functions in a self-contained and independent manner once deployed. It independently verifies or detects a particular type of target and then fires or detonates. ⁸⁴ Such technologies are not new, having been employed for example in the past in mines and booby-traps. ⁸⁵

Autonomous systems differ from automated ones in that they can understand higher-level intent and direction; and ‘from this understanding and its perception of its environment, such a system is able to [take] appropriate action to bring about a desired state’. ⁸⁶ So autonomous systems independently identify and decide to engage targets. They are not pre-programmed to target a specified object or person. It is the software that decides which target to prosecute, how and when.

For the foreseeable future, autonomous attack decision-making is most unlikely to be capable of employment in conformity with targeting law principles and rules, particularly the distinction, discrimination, and proportionality principles and the rules as to precautions in attack. ⁸⁷ At a Chatham House Conference on autonomous weapons systems, there was broad agreement that ‘except in very unique battle spaces (where the likelihood of civilians was nonexistent), deployment of autonomous weapon systems today would not be consistent with the requirements of International Humanitarian Law’. ⁸⁸ Accordingly, a legal review will generally reject the entirely autonomous use of weapon systems employing such technology.

Human Rights Watch has called for a legally binding instrument and national laws banning fully autonomous attack technologies, a notion which it describes as ‘robots that are capable of selecting targets and delivering force without any human input or interaction’ or that, although they operate under the oversight of a human operator who can override the robot’s actions, are subject to such limited supervision that there is no effective human input or interaction. ⁸⁹

A UNIDIR discussion paper argues that there might be a difference in the acceptability of an autonomous but static system that is a ‘last line of defence’ to counter an incoming attack versus a system that employs superhuman decision-making speed to carry out an attack. ⁹⁰ Such a last line of defence system, similar perhaps to Israel’s Iron Dome or the Phalanx system, would be programmed to engage only inbound threats that are by definition military objectives in a defensive operational context in which collateral damage is unlikely to be a prohibitive consideration. Such systems are already in operational use and are likely to be more discriminating than human decisions made in the stressful, high speed, and potentially overwhelming conditions that necessitate the employment of the autonomous/highly automated system.

There has been criticism of the Human Rights Watch report ⁹¹ and its definition of ‘fully autonomous weapons’ which seems to include automated as well as autonomous systems. For the reasons discussed in the previous paragraph, states are likely to consider a prohibition of these technologies premature or inappropriate; so highly automated decision-making, informed certainly by human input before mission commencement, may increasingly become the norm. ⁹²

Improvement of artificial intelligence enabling a weapon system to learn and base its decisions on what it has learned seems likely to be the basis on which autonomy will emerge. Such learning systems might apply lessons learned in the battlespace to develop their own criteria against which to recognise a target; or they may observe and record the pattern of life in the target area and subsequently apply its observations and pre-learned lessons to decide what to attack. Perhaps a future ALI system could detect that a planned attack would no longer comply with the discrimination rule, for example because it detects that hostages have entered the target area.

More realistically, a weapon system might be able to detect whether the previous pattern of life in the target area that it has been observing has changed materially, such that the search for targets by the automated system should not proceed.

States are likely also to consider the ethical issues arising from the acquisition of autonomous weapons, such as whether the decision to initiate the use of lethal force can be legitimately delegated to an automated process. ⁹³ As noted previously, however, it is existing law that should be applied to decide the legal acceptability of such technology in warfare ⁹⁴ and this involves applying the normal weapons law criteria. Thereafter, the weapon review should assess whether the targeting law rules can be complied with despite the absence of a person from target decision-making.

We will start, therefore, by considering the normal weapons law criteria. The superfluous injury and unnecessary suffering principle is likely to be irrelevant, concerned as it is with the nature of the injury caused by the weapon as opposed to the automated or autonomous nature of its targeting decision-making. For similar reasons, the environmental protection rules are unlikely to be relevant to the automated or autonomous aspect of the weapon system.

When applying the indiscriminate weapons rule, the performance of the autonomous or automated target recognition technology, whether during tests or in the course of previous actual hostilities, should be evaluated with care. A weapon system only breaches the indiscriminate weapons rule, however, if it cannot be directed at a specific military objective or if its effects cannot be limited as required by international law and if the result in either case is that the nature of the weapon is to strike military objectives and civilians or civilian objects without distinction. So if attack technology is designed to recognise the particular characteristics of, say, a tank, and if the recognition software performs satisfactorily in tests that realistically reflect the intended circumstances of use the indiscriminate weapons rule is likely to be satisfied. If, however, such software does not perform satisfactorily in such tests and thus the weapon strikes civilian objects and military objectives without distinction, the system will breach the rule. In some cases it may be necessary in the text of the weapon review to draw attention to restricted circumstances in which the weapon system will comply with the indiscriminate weapons principle, ⁹⁵ and to explain what action is required in order to ensure that use of the weapon system does not result in indiscriminate attacks. There are no specific prohibitions or restrictions on the use of autonomous or automated attack technology in customary or treaty law. ⁹⁶

The weapon reviewer should then consider whether the automated or autonomous weapon system is capable of being used in accordance with the targeting rules. This will involve consideration of all targeting rules that seem to be relevant to the particular weapon system. The precautions required of an attacker by article 57 of API illustrate clearly some of the relevant challenges. Article 57(1), to which the reviewer should draw attention, requires that constant care be taken to spare the civilian population, civilians and civilian objects, and this provides the context for what follows.

Those who ‘plan or decide upon an attack’ and thus have obligations under Article 57(2) would seem to include those who prepared the mission, programmed the automated or autonomous software, reviewed available information, prescribed the areas to be searched and the times of such searches, set the target identification criteria for the weapon control software and so on.

The weapon reviewer will need to be satisfied that the characteristics of the weapon system and the arrangements that are being made for its employment are such that the decisions to attack made by the automated or autonomous weapon system apply these provisions, whether by virtue of action taken by the equipment itself or because of what personnel operating the weapon systems, supervising them, or planning the sortie are enabled to do in advance of, or during, the sortie.

Everything ‘feasible’ must be done to fulfil the obligations in Article 57(2)(a)(i) and (ii). If a manned mission would be capable of fulfilling an Article 57 obligation which the automated or autonomous mission cannot fulfil, then the manned mission should be employed, or some other method should be found of achieving the desired military purpose. This may become a difficult issue if an autonomous or automated weapon system cannot be programmed to recognise when its employment would preclude the taking of sub-paragraph (i) or (ii) precautions that, in contrast, could be taken if some other weapon system, such as a manned one, were to be employed. The mere fact that an autonomous or automated system cannot fulfil an obligation does not render the obligation non-feasible if it can be fulfilled using an alternative weapon system traditionally used for the relevant purpose.

The Article 57(2)(a)(i) requirements to do everything feasible to verify that the object of the attack is a military objective and that it is not entitled to special protection ⁹⁷ are likely to be complied with by using algorithm-based technologies, for example, that are found in tests satisfactorily to differentiate between the objects they are programmed to identify and those they are not, and thus between military objects and civilian objects. ⁹⁸

Shifting the focus of attention from targeting objects to targeting persons, the challenge for autonomous or automated target recognition technology under the first element of article 57(2) (a)(i) would be to show that it can satisfactorily distinguish between lawful targets, namely able-bodied combatants and able-bodied civilians directly participating in the hostilities, and persons whom the law protects, such as combatants who are hors de combat, non-combatants, and civilians who do not directly participate.

Article 57(2)(a)(i) also requires that attackers do everything feasible to verify that it is not prohibited by the Protocol to attack the intended targets. Prohibited attacks include those which would breach Article 51(4) (discrimination principle), 51(5)(a) (separate and distinct military objectives treated as one), 51(5)(b) (proportionality), 53 (cultural objects), 54 (objects indispensable to the survival of the civilian population), 35(3) and 55 (protection of the natural environment), 56 (works and installations containing dangerous forces and military objectives in their vicinity), 41 (safeguarding of persons hors de combat), 12 and 15 (protection of medical units and personnel), and 21 to 28 (protection of medical transports).

To the extent that these rules simply prohibit attacks directed at specified objects or persons, the weapon reviewer will be concerned to establish that the automated or autonomous weapon system, in the manner in which it is intended to be used, will detect that a person or object comes within one of these protected categories and will accordingly refrain from attacking it. It remains to be seen whether, for example, software can be developed that distinguishes between an able-bodied combatant and one who comes within article 41 as being hors de combat. ⁹⁹

The precautionary requirements of Article 57, however, go beyond target recognition so an autonomous or automated weapon system must also be able to comply, for example, with the evaluative judgements involved in Article 51(5)(a), in the proportionality assessment referred to in Articles 51(5)(b) and 57, and in the tests in Article 57(2)(a)(ii) and 57(3) of API. Taking Article 57(2) (a)(ii) as an example, the weapon system will have to be able to decide whether an attack should be undertaken using an operator-controlled, automated, or autonomous platform with a view to mini-mising incidental civilian injury and damage. The weapon reviewer will need to be satisfied that the available technology facilitates the making of each of these evaluative assessments.

The difficulty may, however, be overcome if human planners or operators are enabled to make the necessary evaluations and thus take the required precautions ¹⁰⁰ ; but, if there is no such human involvement, the weapon reviewer will need to be satisfied that the weapon system itself can discharge the complex decision-making that has been discussed.

For the foreseeable future, therefore, autonomous or automated attack capabilities can only be used lawfully if required precautions are taken by personnel, probably in advance of the sortie. Appreciating, however, that technology will develop with time and that autonomous or automated weapon systems will tend to be used in conjunction with other support systems, the reviewer’s task will be to determine whether the method of warfare as a whole can comply with the legal rules we have discussed. Current technology requires a person to be in a position to cancel autonomous and some automated attack operations if the need should arise. That person will need to remain sufficiently engaged, suitably located, and appropriately tasked to know what is taking place and, if necessary, to override the system’s attack decisions.