Policy Paper | Autonomous Weapon Systems and Brazil’s Role in Promoting a New International Legal Framework
This policy paper examines the legal, humanitarian, and ethical challenges associated with the development and use of Autonomous Weapon Systems (AWS), emphasizing the need to preserve human control over the use of force. Based on an analysis of international discussions within the framework of the Convention on Certain Conventional Weapons (CCW) and Brazil's position, the document provides recommendations to strengthen Brazil's role in the development of a legally binding international framework that establishes prohibitions and restrictions on AWS and contributes to the protection of civilians and compliance with International Humanitarian Law (IHL).
In this Policy Paper
Introduction
Autonomous Weapon Systems (AWS) are weapons that, once activated, can identify targets and attack them without a person having to decide, in each individual situation, against whom or when force will be used. To do so, they rely on sensors, software and, in some cases, Artificial Intelligence (AI) systems, which enable them to analyse information about their surroundings and recognise characteristics previously programmed to identify potential targets. Based on this analysis, the system itself may select the target and initiate an attack without a specific human decision for each action. This capability raises serious concerns regarding compliance with International Humanitarian Law (IHL), also known as the law of armed conflict, particularly with regard to preserving human judgement in the use of force and applying its fundamental principles. Furthermore, the autonomy of these systems complicates the attribution of responsibility for potential violations, involving different actors, such as States, developers, manufacturers and operators, whose contributions to the decision-making process and the use of force are not always easily distinguishable.
In a context of rapid technological development and the growing integration of Artificial Intelligence (AI) systems into military capabilities, the absence of a specific international legal framework on AWS makes this issue particularly urgent for the protection of civilians and the effective implementation of IHL. The use of these systems entails risks of failures in target identification and classification, arising, among other factors, from technical limitations, inadequate training data and the reproduction of biases embedded in the datasets used. Such limitations may undermine the ability to distinguish military objectives from civilian persons and objects, as well as to assess the specific circumstances of an attack. In response to these challenges, civil society organisations, humanitarian disarmament movements and experts have intensified international advocacy for a legally binding instrument establishing prohibitions and restrictions on the development, production and use of AWS, with a view to preventing humanitarian harm and preserving human control over the use of force.
Against this backdrop, the Seventh Review Conference of the Convention on Certain Conventional Weapons (CCW) constitutes a strategic opportunity to advance the development of an international legal framework on AWS. By bringing together High Contracting Parties to assess the implementation of the Convention and discuss challenges arising from the evolution of conventional weapons, the Conference provides a relevant forum for strengthening international humanitarian protection standards and developing new legal responses to technologies that may undermine compliance with IHL.
The Conference, held every five years, will take place from 16 to 20 November 2026 in Geneva, Switzerland. The Convention is an established instrument of International Humanitarian Law, supplemented by five Protocols addressing, respectively, non-detectable fragments; mines, booby-traps and other devices; incendiary weapons; blinding laser weapons; and explosive remnants of war. The Convention has also become a forum for discussing emerging challenges related to conventional weapons. Since 2013, AWS has been included on its agenda, leading to the establishment of the Informal Meeting of Experts in 2014 and, subsequently, the Group of Governmental Experts (GGE) in 2016. Since then, discussions have sought to deepen understanding of the legal, ethical, operational and humanitarian challenges associated with autonomy in weapon systems, although they have yet to result in the adoption of a specific legally binding instrument.
The urgency of this debate has become even more evident in light of recent incidents involving the use of autonomous technologies in armed conflicts, including the attack reported in Ukraine in July 2026, in which civilian deaths were associated with the use of AWS. These events underscore the need to examine the concrete risks arising from autonomy in the use of force, particularly with regard to civilian protection, accountability for violations and the preservation of human control parameters. At the same time, the expansion of the production and use of armed drones in Latin America, coupled with the potential integration of AI capabilities into these platforms, demonstrates that the issue is not limited to ongoing armed conflicts. It also requires attention at the regional and national levels, considering the risks of proliferation and the development of autonomous military capabilities.
It is in this context that DHESARME - Brazilian Action for Humanitarian Disarmament presents this Policy Paper, with the aim of analysing the humanitarian risks associated with Autonomous Weapon Systems and the international efforts to establish legal limits on their development and use. Based on an analysis of incidents involving these technologies in armed conflicts and the progress of negotiations within the CCW framework, this document examines Brazil’s position in international discussions and the risks associated with the spread of autonomous military technologies in Latin America. It thus seeks to contribute to the debate on the need for a legally binding international instrument establishing prohibitions and restrictions on AWS, ensuring human control over the use of force and strengthening civilian protection, in accordance with the principles and obligations of International Humanitarian Law.
1. The Use of Fully Autonomous Weapon Systems in Ukraine in October 2026 and an Analysis of the Use and Development of These Weapons under International Humanitarian Law
1.1 The Use of Autonomous Weapon Systems in Ukraine in 2026: The Zaporizhzhia Attack and the Challenges to International Humanitarian Law
On 6 July 2026, a Russian attack against a petrol station in the city of Zaporizhzhia, in southeastern Ukraine, resulted in the deaths of three civilians and brought a concrete case into the international debate on the use of Autonomous Weapon Systems. The incident was detailed in an investigation published by The New York Times on 24 August 2026, based on interviews with Ukrainian military officials, drone technology experts and investigators who examined the equipment’s wreckage. The victims were identified as Tetiana Bubynets, aged 19; Oleksiy Svirin, aged 41; and Roman Karpiy, aged 48.
According to the investigation, the equipment used was a Russian Molniya drone fitted with an Nvidia Jetson Orin computing module. Human operators reportedly programmed the equipment to travel to the petrol station. During the final stage, however, the system allegedly autonomously identified and selected the specific target, most likely the propane tanks located at the facility, based on visual patterns previously incorporated into its training. During its approach, the drone failed to avoid a residential building, collided with a wall and exploded before reaching its intended target. The explosion and resulting shrapnel struck people nearby, causing the three deaths. The presence of the computing module and the analysis of data and code recovered from the wreckage supported the assessment by Ukrainian investigators that the equipment had the capability to autonomously identify and select targets.
The case requires an important technical and legal distinction. The equipment’s autonomy did not stem simply from its ability to fly without a pilot or to use artificial intelligence. The relevant element was the delegation to the system of the function of identifying and selecting the target during the final stage of the attack, without a human operator specifically choosing the object to be struck at that moment. In other words, this characteristic reflects the definition of autonomous weapon systems adopted by the International Committee of the Red Cross (ICRC), according to which these weapons, once activated, can select and apply force against targets without human intervention.
The incident also demonstrates that the use of artificial intelligence is not, in itself, sufficient to characterise a system as autonomous. Systems may use algorithms for navigation, image processing or target identification assistance while still requiring a human decision to authorise an attack. In the Zaporizhzhia incident, the most relevant information is that, according to the investigation, the system allegedly autonomously selected the specific target. The case therefore represents a documented example of the use of autonomous target-selection capabilities in an attack that resulted in civilian deaths, although the available information does not allow us to conclude that the system identified the people killed as deliberate targets.
This distinction is essential to avoid conclusions that go beyond the available evidence. The three deaths do not, in themselves, demonstrate that the system was programmed to attack civilians or that it deliberately selected people as targets. The report indicates that the intended target was the propane tanks and that the victims were struck after the equipment collided with the residential structure. Nevertheless, the incident raises questions about the system’s ability to operate predictably, recognise the environment surrounding the target and limit the effects of its operation, particularly in inhabited urban areas.
The attack must also be understood in the context of the expansion of autonomous capabilities in the conflict. On 3 July 2026, Ukrainska Pravda had already reported statements by an adviser to the Ukrainian Ministry of Defence regarding the deployment of autonomous versions of the Molniya drone in the Zaporizhzhia region. The subsequent investigation into the 6 July attack provided additional evidence concerning the equipment’s capabilities and its consequences for civilians. At first glance, the incident demonstrates that the discussion on autonomous weapon systems is no longer confined to hypothetical scenarios of technological development. The incorporation of autonomous target-identification and selection capabilities into the battlefield already raises concrete concerns regarding civilian protection, the predictability of weapons’ effects and the preservation of human responsibility for the use of force.
1.1.1 An Analysis of the Incident under International Humanitarian Law
The use of a technology capable of selecting targets without subsequent human intervention does not preclude the application of the rules governing the conduct of hostilities. IHL remains applicable to the development and use of autonomous weapon systems. However, their technological characteristics may complicate the prior assessment of the legality of their use and the assurance that their effects remain within the established legal limits.
The first point of analysis is the principle of distinction, according to which parties to a conflict must distinguish, in all circumstances, between civilians and combatants, as well as between civilian objects and military objectives. Attacks must be directed exclusively against military objectives, and attacks against civilians are prohibited. The principle also prohibits indiscriminate attacks, including those that are not directed at a specific military objective or whose effects cannot be limited as required by IHL. These obligations are established in Articles 48 and 51 of Additional Protocol I to the Geneva Conventions.
In the case under analysis, the information that the system was allegedly trained to recognise propane tanks and select them as targets raises questions about the criteria used to classify objects and determine their military nature. The visual identification of an object is not equivalent to determining that it constitutes a legitimate military objective. Such a determination requires verifying whether the object effectively contributes to military action and whether its destruction offers a definite military advantage under the circumstances prevailing at the time. A system that identifies objects based on visual patterns may correctly recognise a physical category without necessarily understanding its context, function or military relevance.
This difficulty becomes even more significant in urban environments, where civilian facilities, residential structures and potential military objectives may be located in close proximity. The ability to recognise a particular object does not ensure that the system can adequately assess its surroundings, identify the presence of civilians or adapt its operation to unforeseen situations. Research published by the Stockholm International Peace Research Institute (SIPRI), by Laura Bruun, Marta Bo and Netta Goussac (2023), demonstrates that the compliance of autonomous weapon systems with IHL depends on an assessment of the specific conditions under which these weapons are developed, configured and used. Compliance with humanitarian law cannot be presumed on the basis of an equipment’s stated technical capabilities.
The principle of proportionality must also be considered. Under Article 51(5)(b) of Additional Protocol I, attacks expected to cause incidental loss of civilian life, injury to civilians or damage to civilian objects that would be excessive in relation to the concrete and direct military advantage anticipated are prohibited. In the Zaporizhzhia incident, the deaths of three civilians raise questions regarding the foreseeability of the risks associated with the attack, the choice of means employed and the conditions under which the system was authorised to operate. However, the occurrence of civilian deaths does not, in itself, allow for the conclusion that the principle of proportionality was violated. It is necessary to examine the information available to those responsible for the attack before it was carried out, the anticipated military advantage and the incidental harm that could reasonably have been expected.
The obligation to take precautions in attack, provided for in Article 57 of Additional Protocol I, is also central. Those responsible for planning and carrying out attacks must do everything feasible to verify the military nature of the objectives, choose means and methods that minimise harm to the civilian population, and cancel or suspend the operation when it becomes apparent that the objective is not military or that the expected incidental harm would be excessive. In the case at hand, the navigation failure that caused the drone to collide with a residential building and explode near civilians raises concerns about the system’s ability to operate safely in an urban environment and about the measures adopted to prevent unintended consequences.
The analysis should not be limited to the technical failure that occurred during the attack. It is necessary to consider whether those responsible for the system’s development, programming and use could have foreseen the possibility of navigation errors, recognition failures or explosions at locations other than those initially selected. Furthermore, the predictability, reliability, operational environment and specific conditions of use of the weapons must be examined. The lower the ability to understand and anticipate the system’s behaviour, the more difficult it becomes to ensure that its use complies with humanitarian obligations. Increasing complexity may hinder understanding of how the system operates and the maintenance of effective human control. The formal presence of a human operator, therefore, does not, in itself, guarantee sufficient control over the use of force.
For control to be effective, operators must understand the system’s capabilities and limitations, know the target-selection criteria, assess the environment in which it will be used and have genuine means of intervening or terminating its operation. In the Zaporizhzhia case, the autonomous selection of the specific target demonstrates the reduction of human intervention precisely at a decisive stage of the attack. Human involvement in the initial stages, such as programming a route or defining an area of operation, does not necessarily mean that sufficient control exists over the final decision to use force.
Another relevant aspect is legal responsibility. The autonomy of a system does not eliminate the responsibility of States and individuals involved in its development, acquisition, programming, authorisation and use. However, the greater the system’s complexity and the lower the transparency of its decision-making processes, the more difficult it may become to reconstruct the circumstances of an attack and identify responsibility for potential violations.
The investigation into the Zaporizhzhia case, based on the analysis of components and code recovered from the drone, demonstrates the importance of preserving technical evidence that enables an understanding of the weapon’s functioning and the decisions embedded in its target-selection process.
For this reason, the legal review of new weapons, provided for in Article 36 of Additional Protocol I, assumes particular importance. States must determine, during the study, development, acquisition or adoption of a new weapon, whether its use would be prohibited in some or all circumstances under applicable international law. In the case of autonomous systems, this review must consider the capabilities declared by the manufacturer, as well as the system’s behaviour in different environments, the reliability of its identification mechanisms, the limits of its predictability and the possibility of ensuring effective human control during its use.
1.2 Digital Dehumanisation, Human Dignity and the Humanitarian Risks of Autonomy
Beyond the challenges involved in applying IHL, the development of autonomous weapon systems raises concerns relating to human dignity and the ethical limits of delegating lethal decisions to machines. One of the main issues identified by civil society organisations is the possibility of digital dehumanisation, a process through which individuals come to be represented and classified based on data, visual patterns and computational profiles, replacing consideration of their specific circumstances.
In a report published in 2025, Human Rights Watch and the International Human Rights Clinic at Harvard Law School, entitled A Hazard to Human Rights: Autonomous Weapons Systems and Digital Decision-Making, examine the implications of these systems for the right to life, human dignity, non-discrimination and access to remedies. The organisations argue that the use of algorithms to identify and select targets may reduce individuals to sets of data and computational characteristics, subjecting decisions over life and death to processes devoid of human moral judgement.
This concern is further explored by Automated Decision Research in its publication on autonomous weapons and digital dehumanisation. The organisation links the growing use of automated systems to the reduction of people to computational patterns and categories, highlighting the risks of replacing contextualised assessments with algorithmic classifications. The problem is not limited to the possibility of identification errors. Even when technical results are considered accurate, the question remains of whether decisions involving the assessment of human circumstances and the use of lethal force should be delegated to computational systems.
The use of data to identify individuals, recognise behaviours or establish target profiles may also reproduce discrimination and biases embedded in the datasets used to train these systems. As a result, certain physical, behavioural or contextual characteristics may be erroneously associated with situations involving threats. The difficulty of understanding the criteria used by an algorithm may, in turn, hinder the identification of errors and the possibility of challenging decisions.
Although the Zaporizhzhia incident reportedly involved the selection of an object as a target, rather than the autonomous identification of individuals, its analysis provides an opportunity to discuss the risks posed by systems that classify elements of their environment based on predefined visual patterns. When this logic is applied to the identification of individuals, the consequences may be even more severe, particularly in environments where distinguishing civilians from combatants requires contextual assessments.
The humanitarian debate is not limited to the errors these systems may make or to the difficulties involved in applying existing legal rules. It also concerns human dignity itself and the limits of transferring to machines decisions that may determine whether people live or die. In this regard, the Martens Clause, by recognising the importance of the principles of humanity and the dictates of public conscience, provides an important reference point for reflecting on the limits of such delegation. After all, even if a system operates in accordance with the parameters for which it was programmed, this does not mean that every decision it may make is legally or ethically acceptable.
1.3 The Need for Specific Legal Limits from a Humanitarian Disarmament Perspective
The Zaporizhzhia attack demonstrates that the risks associated with autonomous weapon systems are not merely hypothetical. Autonomous target selection has already been linked to a documented incident involving civilian deaths, showing that the incorporation of artificial intelligence into identification and engagement functions can produce concrete humanitarian consequences. Although the available information does not allow the conclusion that the system deliberately selected civilians as targets, the incident reveals the risks arising from the combination of automated selection, operational unpredictability and reduced human intervention.
IHL remains binding in armed conflicts. Nevertheless, the characteristics of autonomous weapon systems raise questions that are not always adequately addressed by existing general rules. The international debate must advance towards defining which forms of autonomy may be permitted and which exceed the limits necessary to preserve human control and the dignity of people affected by conflicts.
To understand this discussion, it is important to consider the trajectory of humanitarian disarmament. The development of new international instruments in this field has been shaped by the interaction between evidence of the impacts of weapons, civil society mobilisation and State participation. The processes that led to the adoption of the Convention on the Prohibition of Anti-Personnel Mines in 1997 and the Convention on Cluster Munitions in 2008 are examples of this dynamic, bringing together humanitarian organisations, researchers, survivors and transnational movements around common objectives.
These experiences show that humanitarian disarmament involves work that extends beyond the adoption of prohibitions and restrictions. Treaty implementation depends on independent monitoring, prevention, victim assistance, the clearance of explosive remnants, transparency and the monitoring of States’ commitments. In this context, the Landmine Monitor and Cluster Munition Monitor reports are important references for understanding how civil society generates information, monitors treaty implementation and calls for compliance with international obligations.
The work of Action on Armed Violence (AOAV), particularly through theExplosive Violence Monitor, provides an additional contribution to this debate. Its findings make it possible to monitor the effects of explosive weapons on civilians, with particular attention to attacks in populated areas. Although it does not specifically address autonomous weapon systems, this type of monitoring helps situate the discussion of new military technologies within its humanitarian context, taking into account the concrete impacts of their development and use.
In the field of autonomous weapons, organisations such as Human Rights Watch, Article 36 and PAX have contributed to deepening the debate on the limits of human control. Their work indicates that formal oversight by an operator is insufficient to guarantee such control. Those responsible for employing the systems must understand their capabilities and limitations, be able to assess the circumstances of their use and have genuine opportunities to intervene. Technological complexity, algorithmic opacity and the risks of discrimination must also be taken into account, as they directly affect the protection of life and human dignity.
Research by the Stockholm International Peace Research Institute (SIPRI) and the United Nations Institute for Disarmament Research (UNIDIR) complements these contributions by examining the application of IHL, the limits of existing legal interpretations and States’ positions in the multilateral debate.
As pesquisas do Stockholm International Peace Research Institute (SIPRI) e do United Nations Institute for Disarmament Research (UNIDIR) complementam essas contribuições ao examinar a aplicação do DIH, os limites das interpretações jurídicas existentes e as posições dos Estados no debate multilateral.
Civil society mobilisation has also been important in promoting negotiations on a specific legal framework. The Stop Killer Robots campaign brings together organisations and actors from different regions that advocate for a legally binding international instrument prohibiting autonomous weapon systems that operate without meaningful human control and imposing mandatory restrictions on other forms of autonomy in weapons. This work highlights an issue that goes beyond the assessment of systems’ technical capabilities: the definition of limits that should guide the automation of decisions related to the use of force.
The Zaporizhzhia incident reinforces the relevance of these concerns. The circumstances of the attack demonstrate the challenges associated with the predictability of systems’ behaviour, target identification and civilian protection. Although the application of IHL is indispensable for assessing the use of these technologies, the difficulties involving distinction, proportionality, precautions and accountability also demonstrate the importance of establishing more specific legal limits.
The humanitarian disarmament perspective makes it possible to advance this discussion. It is necessary to define which functions related to the use of force should not be delegated to autonomous systems, which applications must be prohibited and which restrictions are necessary to prevent unacceptable humanitarian consequences. Negotiating a legally binding international instrument therefore represents a means of establishing these limits preventively, reducing risks to civilians and ensuring that decisions on the use of force remain subject to human judgement and accountability.
2. The Discussion of Autonomous Weapon Systems within the Framework of the Convention on Certain Conventional Weapons and the Urgency of a Legal Framework
Discussions on autonomous weapon systems within the United Nations (UN) framework have taken place primarily under the Convention on Certain Conventional Weapons (CCW), which has served as the main multilateral forum bringing together State delegations and experts since 2014 to examine the various issues related to autonomous weapons. It is worth noting that, between 2010 and 2013, before discussions became concentrated within the CCW, autonomous weapons had already been addressed by civil society organisations and UN Special Rapporteurs. Christof Heyns’s report to the Human Rights Council referred to them as “lethal autonomous robotics”, warning of the urgency of action and recommending national moratoria (Blanchard & Goussac, 2025; Heyns, 2013).
Initially, discussions on AWS within the CCW took place through informal meetings between 2014 and 2016, consisting of presentations by civil society experts addressing legal, ethical, military, technical and security issues. In general, there was agreement that international law applies to the development and use of AWS, although disagreements emerged regarding their technical ability to comply with IHL principles relating to distinction, proportionality and precautions in the use of force (Boulanin, 2016).
In 2016, during the Fifth CCW Review Conference, the High Contracting Parties decided to establish the Group of Governmental Experts on Lethal Autonomous Weapon Systems (GGE), open to States Parties, international organisations and civil society organisations. The Group would meet the following year to examine in greater detail the issues associated with AWS in relation to the objectives of the Convention (Blanchard & Goussac, 2025; Davis & Verbruggen, 2018). From its first meeting in 2017, the GGE’s initial work focused on four areas: (a) the characterisation of autonomous weapons; (b) possible legal and ethical challenges; (c) military implications; and (d) aspects of human-machine interaction in the development and use of AWS (Blanchard & Goussac, 2025). During this initial period, attempts to adopt a working definition did not advance due to disagreements over its scope and formulation (Davis & Verbruggen, 2018).
An important milestone occurred in 2019, when the GGE adopted a list of 11 guiding principles to steer its future work. Broadly speaking, the principles affirmed that IHL continues to apply fully to all weapon systems, including AWS; that human responsibility for decisions concerning the use of these systems must be maintained throughout their entire life cycle, since accountability cannot be transferred to machines; and that human-machine interaction must ensure compliance with applicable international law, taking into account factors such as the operational context and the characteristics of the system. They also established that new weapons must be assessed for their compliance with international law, that risks such as proliferation and acquisition by terrorist groups must be considered, and that risk assessments and mitigation measures should be incorporated into the design, development, testing and deployment cycle (Group of Governmental Experts on Lethal Autonomous Weapons Systems [GGE], 2019).
In 2021, the Sixth Review Conference failed to reach an agreement to commence negotiations and adopted a mandate to consider proposals and develop possible measures concerning AWS. According to Human Rights Watch (2021), Russia, the United States, India and Israel were primarily responsible for preventing the approval of a negotiating mandate, given the consensus-based decision-making process within the CCW. In 2023, the Meeting of the High Contracting Parties decided to continue the GGE’s work, mandating it to formulate, by consensus, a set of elements for an instrument, without prejudging its nature, and to submit a report to the Seventh CCW Review Conference (GGE, 2026). Between 2024 and 2026, under the chairmanship of Ambassador Robert in den Bosch of the Netherlands, the Group’s work came to be based on the rolling text proposed by the Chair (Blanchard & Goussac, 2025; GGE, 2025).
The GGE’s final session, held from 31 August to 4 September 2026, was the last under its mandate and would determine what would be submitted to the Seventh Review Conference. On the first day, more than 70 States declared their support for negotiations based on the elements developed by the GGE (Varella, 2026a). A group of 47 countries, in a joint statement presented by Brazil, reiterated that the June 2026 rolling text, included in the draft report, was sufficient to fulfil the GGE’s mandate and serve as a basis for negotiations within the CCW (Brasil, 2026; Varella, 2026b). Throughout the week, however, several informal meetings were held exclusively among the High Contracting Parties, excluding observers such as regional and international organisations, NGOs, civil society and academia (Varella, 2026b). The Chair presented a revised version of the text on Thursday and another on Friday, when the discussion was conducted entirely in an informal format, with a brief return to the formal format for an interpretative reading of paragraph 25 concerning the characterisation of autonomous weapons. The final report was adopted by consensus on 4 September 2026 (GGE, 2026; Varella, 2026b).
The report brings together a set of elements for a future instrument, including, among others, a characterisation defining lethal AWS as systems capable of identifying, selecting and engaging targets in a manner resulting in loss of life, without the intervention of a human operator (para. 25); the reaffirmation of the full applicability of IHL (paras. 27–29); the requirement for human control and judgement to ensure compliance with IHL (paras. 34–35); and provisions on legal reviews, biases and accountability (paras. 36–41). The report does not, however, call for the negotiation of an instrument. Instead, it recommends that the States Parties, at the Seventh Review Conference, “consider and endorse” the elements and decide by consensus on the modalities for future work (para. 45). Thus, an agreed text now exists, and it is up to the States Parties to decide whether to transform it into an international legal instrument (GGE, 2026).
The final report’s content revisits some of the 2019 guiding principles: it maintains that human accountability cannot be transferred to machines (paras. 40–41), that systems must operate under a responsible chain of command and control (para. 35), and that risks such as proliferation and acquisition by terrorist groups must be considered (para. 37). Obligations concerning use are framed around applicable international law in armed conflict (paras. 33, 34 and 43), wording that coincides with the proposal submitted by the United States and supported by Australia, Canada, India, Israel and the Republic of Korea. Conversely, Austria, Brazil, Chile, Colombia, Ireland and several other delegations supported the reference to applicable international law, particularly IHL (GGE, 2026; Varella, 2026b).
However, some elements were left out of the final text. According to Laura Varella of Reaching Critical Will, the informal consultations excluded paragraphs on ethical considerations and the legal review of systems following modifications, as well as provisions concerning concepts such as reliability, predictability and traceability, among other topics. This resulted in content described by the author as “significantly diluted” (Varella, 2026a). For Verity Coyle (2026) of Human Rights Watch, certain States, including Russia and the United States, weakened the text by insisting on changes to widely supported provisions. Furthermore, the final text does not provide for the prohibition of anti-personnel autonomous weapons, which had been advocated by some delegations and observers during the first 2026 session (Chairperson of the GGE, 2026).
Support for commencing negotiations based on the elements is substantial: 76 States, 70 of them among the Convention’s 128 States Parties, representing a majority (Automated Decision Research, 2026a). However, given that CCW decisions are made by consensus, opposition from a minority of States could block substantive progress under a potential new negotiating mandate. At present, taking into account other multilateral forums and not only the CCW, twelve States oppose the creation of a legally binding instrument on autonomous weapons: Australia, Belarus, the Democratic People’s Republic of Korea, Estonia, India, Israel, Japan, Poland, the Republic of Korea, Russia, the United Kingdom and the United States. In contrast, 130 States support it, while 53 have not stated their position (Automated Decision Research, 2026b).
As discussed in the previous section, IHL already applies to autonomous weapons, but it does not address all the questions they raise. It is in this context that a specific legal framework for AWS becomes necessary. For example, the GGE’s final report addresses human-machine interaction in paragraphs 34–35, but formulates the measures concisely and with qualifications such as “as appropriate” and “to the extent possible or feasible”, while some of the other provisions are framed as recommendations (paras. 38, 39, 42 and 44) (GGE, 2026). In this regard, in August, before the final meetings, the International Committee of the Red Cross (ICRC) assessed that the set of elements in the July draft report provided a sufficient basis for commencing negotiations. However, it maintained that a binding instrument would need to include more specific prohibitions and restrictions, establishing obligations beyond those already provided for under IHL. Among these, it highlighted the imperative of prohibiting autonomous weapons designed or used to directly target human beings, also referred to as anti-personnel autonomous weapons. According to the ICRC, the absence of such a prohibition in the draft significantly undermines the effectiveness of the set of elements (International Committee of the Red Cross [ICRC], 2026).
The GGE’s mandate has been completed: the Group formulated, by consensus, a set of elements for an instrument, without prejudging its nature. The Seventh CCW Review Conference, scheduled for 16–20 November, will decide the future direction of the work, potentially approving a negotiating mandate, extending the discussions or allowing the process to expire. A possible binding instrument could take the form of a new Protocol to the Convention or an independent treaty (Blanchard & Goussac, 2025). After more than a decade of discussions within the UN, the CCW and GGE processes have fostered progress in understanding AWS and achieved a consensual outcome. The commencement of negotiations, however, also depends on a decision requiring consensus, making the Review Conference in November a critical moment that will determine the direction of international efforts to prohibit and regulate autonomous weapons in the coming years.
3. Progress in Discussions on Autonomous Weapon Systems and Militarised Artificial Intelligence at the National and Regional Levels
In public perception, the issue of AWS is often associated with a collective impression that tends to distance the problem of autonomy in weapon systems, portraying it either as a future, even “fictional”, concern or as an issue unrelated to Brazilian and Latin American realities. However, the reduction of human intervention in the decision-making processes of armed systems and the integration of AI are becoming increasingly relevant to the regional context, with discussions ranging from advances in the national regulation of these technologies to the use of armed drones by criminal groups.
At the international level, Brazil has advocated for the adoption of a legally binding instrument on AWS for more than a decade, with the preservation of human control over the use of force as its central principle. This position has been developed primarily within the framework of the Convention on Certain Conventional Weapons, in whose discussions the country participated through the informal expert meetings initiated in 2014 and supported, in 2016, the establishment of the Group of Governmental Experts on Lethal Autonomous Weapon Systems.
Brazil’s engagement gained particular significance when the country assumed the chairmanship of the GGE-LAWS in 2022 and 2023. Subsequently, in September 2025, Brazil, together with Ireland, Norway and Switzerland, coordinated a declaration supporting the rolling text and the opening of negotiations. Endorsed by 42 High Contracting Parties to the CCW, the declaration reaffirmed that:
[...] the rolling text of 5 June, now included in subsection A of the draft substantive report of the GGE, remains coherent and well-structured. We therefore continue to consider that it constitutes a sufficient basis to fulfil the mandate of this GGE and to serve as a foundation for negotiations within the CCW on an instrument concerning lethal autonomous weapon systems. (Author’s own translation)
In November 2025, Brazil once again expressed its support for the development of a legally binding instrument and highlighted the Seventh CCW Review Conference, scheduled for November 2026, as a decisive opportunity to advance negotiations.
Brazil’s position is also supported by regional coordination initiatives. In February 2023, Brazil joined the group of 33 Latin American and Caribbean States that adopted the Belém Communiqué. Among its main points, the communiqué highlights the need to preserve meaningful human control over the use of force and to ensure mechanisms for individual and State accountability. The document states:
As new weapons technologies continue to be developed and used in armed conflicts, new prohibitions and regulations guided by international law, including international humanitarian law and its principles, based on the principles of humanity, the dictates of public conscience and ethical perspectives, are urgently needed. In accordance with international law, including the purposes and principles enshrined in the Charter of the United Nations, international human rights law and international humanitarian law, the international community is called upon to respond to these threats through the development and strengthening of the international legal framework, in order to prevent the effects of displacement, injury and death of civilians, as well as the destruction of civilian objects. Maintaining meaningful human control is essential to prevent further dehumanisation of warfare, as well as to ensure individual accountability and State responsibility. (Author’s own translation)
At the national level, the debate on the regulation of Artificial Intelligence has advanced through Bill No. 2,338/2023, authored by Senator Rodrigo Pacheco, which proposes the establishment of a regulatory framework for the development and use of AI systems in Brazil. The proposal establishes rights for individuals affected by these systems, provides for their classification according to risk levels, defines governance obligations for developers and agents responsible for their use, and establishes institutional mechanisms for regulation and oversight. The measures proposed include prohibiting applications deemed to pose excessive risk, requiring impact assessments and documentation for high-risk systems, and providing for human supervision mechanisms.
Approved by the Senate Plenary on 10 December 2024, the text classifies AI systems used in autonomous weapon systems as posing excessive risk under Article 13, prohibiting their development, implementation and use. This wording resulted from a dispute during the bill’s legislative process. The preliminary substitute text presented in 2024 had permitted the use of AWS under certain circumstances, a provision criticised by civil society organisations such as Coalizão Direitos na Rede due to the possibility of employing these systems in public security. The prohibition was reinstated in the final version of the text.
Referred to the Chamber of Deputies in March 2025, the bill is being considered under a priority procedure by the Special Committee on Artificial Intelligence, chaired by Deputy Luisa Canziani, with Deputy Aguinaldo Ribeiro serving as rapporteur. At the time covered by this assessment, the proposal was awaiting the rapporteur’s opinion and had more than 37 related bills attached to it. The rapporteur had also indicated that the Committee’s vote would take place only after the electoral period. Since the proposal originated in the Senate, if the Chamber of Deputies approves it without amendments, the text will proceed directly to presidential approval or veto. If amendments are introduced, the text will return to the Senate, which will deliberate on the changes made by the deputies and may either retain them or reinstate the original wording. Only then will it be submitted for presidential approval or veto.
While awaiting deliberation by the Chamber of Deputies, Bill No. 2,338/2023 remains Brazil’s main initiative to address the risks associated with artificial intelligence. Its scope extends beyond AWS, regulating the use of biometric identification systems, the processing of sensitive data and information security, among other applications.
The existence of a proposed prohibition in the national legal system, however, does not replace the need for international rules on the subject. Domestic regulation, although relevant, is insufficient to establish common obligations among States or to independently prevent the development, transfer and use of AWS in other countries. For this reason, Brazil’s engagement in the multilateral system, particularly within the CCW, must keep pace with domestic legislative developments. Coherence between these two areas of action is essential for Brazil to strengthen its diplomatic position and contribute to the development of an international legal framework oriented towards civilian protection and respect for human rights and IHL.
3.1 Armed Drones in Latin America: Rising Risks amid the Trend towards Integrating Artificial Intelligence into Military Systems
In Latin America and the Caribbean, the debate on the risks associated with Autonomous Weapon Systems must also take into account the region’s political, social and institutional conditions. Social inequality, poverty, corruption, institutional fragility and the exclusion of historically marginalised groups are relevant factors. In this context, the potential incorporation of AWS into military and security capabilities may deepen existing vulnerabilities, increase the lethality of violence and hinder civilian protection, thereby undermining regional peace and security. The fight against drug trafficking and other forms of organised crime is a particularly relevant dimension of this discussion, given that growing access to autonomous military technologies could expand criminal organisations’ surveillance and territorial control capabilities, as well as intensify disputes over strategic routes and areas.
In Brazil, the use of armed drones by criminal organisations already represents a concrete public security challenge. In Rio de Janeiro, for example, these devices have been used by criminal organisations to monitor police movements, surveil communities, transport weapons and deploy explosives. According to a Civil Police assessment, at least 17 attacks involving these devices have been recorded since 2025, 12 of which occurred in 2026. Criminals have also resorted to homemade modifications to turn commercial drones into weapons, using low-cost devices to release explosives remotely. In 2026 alone, 31 drones were seized from individuals suspected of belonging to criminal organisations, demonstrating the incorporation of these technologies into urban violence dynamics.
Although the use of drones does not necessarily entail autonomy in target selection and engagement, this scenario illustrates the challenges of control and prevention associated with the proliferation of military technologies. In the absence of international rules capable of establishing clear limits on the development and use of autonomous weapon systems, the expansion of these technologies may exacerbate existing risks, facilitating their adoption by State and non-State actors and increasing the difficulties involved in civilian protection and accountability for the harm caused.
In addition to the risks associated with organised crime, it is also necessary to consider the possible consequences of employing autonomous weapon systems in State security and repression operations. In Latin America, racialised groups, low-income populations, activists and other historically marginalised sectors are already exposed to different forms of institutional violence. The introduction of systems capable of identifying, selecting and attacking targets without direct human intervention may exacerbate these vulnerabilities, particularly when identification processes are based on data that reproduce discriminatory patterns. The absence of adequate human control and accountability mechanisms may also hinder the investigation of violations and undermine the protection of human rights.
In light of these challenges, regional diplomatic engagement constitutes an important dimension of the debate. According to a survey conducted by Stop Killer Robots, based on official State statements during meetings of the United Nations General Assembly and the Convention on Certain Conventional Weapons, all Latin American countries have expressed support for the creation of an international legal instrument regulating autonomous weapon systems. This position indicates regional political convergence around the need to advance the development of international rules to address the risks posed by these technologies.
Figure - Latin American countries supporting negotiations on a legally binding instrument on autonomous weapons, updated in August 2026.
This political alignment therefore provides a basis for expanding regional cooperation and strengthening Latin American participation in multilateral negotiations. Drawing on their experiences, countries in the region can contribute to the development of rules that prioritise civilian protection, preserve human control over the use of force and establish accountability mechanisms. In this way, joint action can strengthen the region’s contribution to the international governance of military technologies and reaffirm its commitment to human rights, IHL and the prevention of violence.
Conclusion and Recommendations
The risks associated with Autonomous Weapon Systems can no longer be treated as a distant concern or one limited to future scenarios of armed conflict. Recent episodes involving the use of military technologies with varying degrees of autonomy, including the Zaporizhzhia case, demonstrate the need to deepen the discussion on the limits of the use of force and the protection of civilians. Although IHL establishes principles and obligations applicable to the conduct of hostilities, its application alone does not replace the need for specific rules establishing clear limits on the development and use of systems capable of selecting and attacking targets without human intervention. The absence of a specific legal framework maintains significant gaps concerning harm prevention, human control and accountability for potential violations.
In this context, the Seventh Review Conference of the Convention on Certain Conventional Weapons represents a decisive opportunity to transform more than a decade of diplomatic discussions into concrete normative advances. Brazil should use this forum to advocate for the establishment of a mandate to negotiate a legally binding international instrument, with provisions prohibiting weapon systems that operate without meaningful human control and establishing strict restrictions on those whose functioning presents risks incompatible with IHL. This instrument should also provide for mechanisms of human supervision, risk assessment, legal review and accountability, covering the different stages of the development, production, transfer and use of these weapons.
At the national level, it is essential for Brazil to maintain coherence between its diplomatic engagement and domestic legislative developments. In this regard, preserving the prohibition set out in Article 13 of Bill No. 2,338/2023 represents an important measure for establishing limits on the development and use of AI systems employed in autonomous weapons. The country should ensure that this approach is maintained throughout the legislative process and that future defence, public security and technological development policies incorporate safeguards consistent with the protection of human rights and its international obligations. Domestic regulation, however, should not be understood as a sufficient response: Brazil must continue advocating for international rules capable of establishing common obligations and preventing the absence of regulation in other States from undermining the effectiveness of these measures.
At the regional level, Brazil is well positioned to play a significant role in coordinating Latin American and Caribbean countries around this agenda. The position expressed in the Belém Communiqué and regional support for the creation of a legally binding instrument provide a political basis for expanding diplomatic cooperation and strengthening joint participation in multilateral negotiations. It is therefore recommended that the country promote regional consultations, encourage the development of common positions and expand dialogue on the risks associated with the proliferation of military technologies, also taking into account the challenges related to the use of armed drones by organised crime. This discussion should distinguish the use of remotely operated drones from genuinely autonomous systems, while recognising the risks arising from technological developments and their potential integration into autonomous military capabilities.
Finally, organised civil society should be recognised as an indispensable part of this process. Humanitarian organisations, researchers, experts and disarmament movements play a fundamental role in producing knowledge, identifying risks and monitoring the positions adopted by States. In this regard, it is recommended that Brazil expand opportunities for dialogue and participation by these actors in the formulation of its national positions and in international discussions, ensuring transparency and continuity in the diplomatic commitments undertaken.
The elements gathered in this document demonstrate the complexity of the issue of AWS and the urgency of creating a legally binding international framework for these weapons. The protection of human life, the preservation of dignity and accountability for violations must not be subordinated to the pursuit of operational advantages. Brazil has the opportunity and the responsibility to contribute to this process by strengthening its national, regional and multilateral engagement in support of rules that keep decisions over life and death under human control. The Seventh CCW Review Conference should therefore represent a concrete step in this direction.
DHESARME’s Recommendations to the Brazilian Government
1. Within the CCW: Advocate for a mandate to negotiate a legally binding instrument on AWS during the Seventh Review Conference.
2. In national legislation: Preserve the prohibition of AWS in Bill No. 2,338/2023 and ensure its compatibility with Brazil’s diplomatic positions.
3. In Latin America and the Caribbean: Lead diplomatic consultations to consolidate a common regional position and expand countries’ participation in negotiations.
4. In regional security: Promote cooperation to monitor the proliferation of military technologies and prevent their use against civilian populations.
5. In social participation: Institutionalise dialogue with civil society organisations, experts and humanitarian organisations in the formulation and monitoring of Brazil’s policy on AWS.
Dhesarme - Brazilian Action for Humanitarian Disarmament
12 November 2026
Technical Team
Dhesarme Team: Fernando Fiala, Hevelyn Ghizzi, Júlia Marcon and Larissa Parcianelo.
Dhesarme Volunteer Network: Gabriel Debatin Coutinho.
References
Action on Armed Violence. (2026). Explosive violence monitor 2025. https://aoav.org.uk/2026/explosive-violence-monitor-2025/
Agência Câmara. (2025, 20 de maio). Comissão que vai analisar projeto sobre inteligência artificial será presidida por Luisa Canziani; Aguinaldo Ribeiro será relator. https://www.camara.leg.br/noticias/1160383-comissao-que-vai-analisar-projeto-sobre-inteligencia-artificial-sera-presidida-por-luisa-canziani-aguinaldo-ribeiro-sera-relator/
Automated Decision Research. (2022). Autonomous weapons and digital dehumanisation: A short explainer paper. https://automatedresearch.org/news/report/autonomous-weapons-and-digital-dehumanisation-a-short-explainer-paper
Automated Decision Research. (2026a, 4 de setembro). Over 70 states now support elements work at CCW GGE as a basis for negotiations. https://automatedresearch.org/news/over-70-states-support-rolling-text-as-basis-for-negotiations/
Automated Decision Research. (2026b). State positions. https://automatedresearch.org/state-positions/
Automated Decision Research. (2026c). State positions: Latin America. https://automatedresearch.org/state-positions/?_state_position_region=latin-america
Blanchard, A., & Goussac, N. (2025). Towards multilateral policy on autonomous weapon systems. Stockholm International Peace Research Institute. https://doi.org/10.55163/VLLZ7787
Boulanin, V. (2016). Mapping the debate on LAWS at the CCW: Taking stock and moving forward (EU Non-Proliferation Paper No. 49). EU Non-Proliferation Consortium. https://www.sipri.org/publications/2016/eu-non-proliferation-papers/mapping-debate-laws-ccw-taking-stock-and-moving-forward
Brasil. (2024). Projeto de Lei nº 2.338, de 2023: Dispõe sobre o uso da Inteligência Artificial. Senado Federal. https://www25.senado.leg.br/web/atividade/materias/-/materia/157233#tramitacao_10494842
Brasil. (2026). Joint statement to the Group of Governmental Experts on Lethal Autonomous Weapon Systems August–September 2026 session. https://reachingcriticalwill.org/images/documents/Disarmament-fora/ccw/2026/gge/statements/31Aug_Joint_Brazil.pdf
Bruun, L., Bo, M., & Goussac, N. (2023). Compliance with international humanitarian law in the development and use of autonomous weapon systems: What does IHL permit, prohibit and require? Stockholm International Peace Research Institute. https://www.sipri.org/publications/2023/policy-reports/compliance-international-humanitarian-law-development-and-use-autonomous-weapon-systems-what-does
Chairperson of the Group of Governmental Experts on Emerging Technologies in the Area of Lethal Autonomous Weapons Systems. (2026). Chair's summary – First 2026 session on the GGE on LAWS (CCW/GGE.1/2026/WP.2). United Nations. https://docs-library.unoda.org/Convention_on_Certain_Conventional_Weapons_-Group_of_Governmental_Experts_on_Lethal_Autonomous_Weapons_Systems_(2026)/CCW-GGE.1-2026-WP.2.pdf
Coalizão Direitos na Rede. (2024, 18 de maio). Por uma legislação de IA centrada em direitos: Comentários ao texto preliminar substitutivo do 2338/2023 apresentado pelo Senador Eduardo Gomes. https://direitosnarede.org.br/2024/05/18/por-uma-legislacao-de-ia-centrada-em-direitos-comentarios-ao-texto-preliminar-substitutivo-do-2338-2023-apresentado-pelo-senador-eduardo-gomes/
Costa Rica. Ministerio de Relaciones Exteriores y Culto. (2023, 23–24 de fevereiro). Communiqué of the Latin American and the Caribbean Conference of Social and Humanitarian Impact of Autonomous Weapons. https://www.rree.go.cr/files/includes/files.php?id=2261&tipo=documentos
Coyle, V. (2026, 7 de setembro). UN talks on killer robots end with calls for negotiations growing. Human Rights Watch. https://www.hrw.org/news/2026/09/07/un-talks-on-killer-robots-end-with-calls-for-negotiations-growing
Davis, I., & Verbruggen, M. (2018). The Convention on Certain Conventional Weapons. In SIPRI Yearbook 2018. Stockholm International Peace Research Institute. https://www.sipri.org/sites/default/files/SIPRIYB18c09sI.pdf
De Souza, R. (2026, 21 de setembro). Ataque aéreo: Drones que lançam bombas mudam padrão dos confrontos do crime organizado no Rio. O Globo. https://oglobo.globo.com/rio/noticia/2026/09/21/ataque-aereo-drones-que-lancam-bombas-mudam-padrao-dos-confrontos-do-crime-organizado-no-rio.ghtml
Group of Governmental Experts on Emerging Technologies in the Area of Lethal Autonomous Weapons Systems. (2019). Report of the 2019 session (CCW/GGE.1/2019/3). United Nations. https://documents.unoda.org/wp-content/uploads/2020/09/CCW_GGE.1_2019_3_E.pdf
Group of Governmental Experts on Emerging Technologies in the Area of Lethal Autonomous Weapons Systems. (2025). Rolling text, status date: 12 May 2025. United Nations. https://docs-library.unoda.org/Convention_on_Certain_Conventional_Weapons_-Group_of_Governmental_Experts_on_Lethal_Autonomous_Weapons_Systems_(2025)/Revised_rolling_text_as_of_12_May_2025.pdf
Group of Governmental Experts on Emerging Technologies in the Area of Lethal Autonomous Weapons Systems. (2026). Report of the 2024–2025–2026 sessions (CCW/GGE.1/2026/3) [Advance copy]. United Nations. https://docs-library.unoda.org/Convention_on_Certain_Conventional_Weapons_-Group_of_Governmental_Experts_on_Lethal_Autonomous_Weapons_Systems_(2026)/CCW-GGE.1-2026-3-Advance_copy.pdf
Heyns, C. (2013). Report of the Special Rapporteur on extrajudicial, summary or arbitrary executions (A/HRC/23/47). United Nations Human Rights Council. https://www.ohchr.org/sites/default/files/Documents/HRBodies/HRCouncil/RegularSession/Session23/A-HRC-23-47_en.pdf
Human Rights Watch. (2021, 19 de dezembro). Killer robots: Military powers stymie ban. https://www.hrw.org/news/2021/12/19/killer-robots-military-powers-stymie-ban
Human Rights Watch, & International Human Rights Clinic at Harvard Law School. (2025, 28 de abril). A hazard to human rights: Autonomous weapons systems and digital decision-making. https://www.hrw.org/report/2025/04/28/a-hazard-to-human-rights/autonomous-weapons-systems-and-digital-decision-making
International Committee of the Red Cross. (1977, 8 de junho). Protocol additional to the Geneva Conventions of 12 August 1949, and relating to the protection of victims of international armed conflicts (Protocol I). https://ihl-databases.icrc.org/en/ihl-treaties/api-1977
International Committee of the Red Cross. (2026). Working paper submitted to the CCW Group of Governmental Experts on emerging technologies in the area of Lethal Autonomous Weapons Systems. https://www.icrc.org/sites/default/files/2026-08/ICRC_working_paper_GGE_LAWS_August-2026-ICRC.pdf
Kramer, A. E. (2026, 24 de agosto). A drone killed three Ukrainians. It was guided entirely by A.I. The New York Times. https://www.nytimes.com/2026/08/24/world/europe/russia-drones-autonomous-ai-kill-ukraine-war.html
Landmine and Cluster Munition Monitor. (2025). Landmine monitor 2025. https://the-monitor.org/online-reader/landmine-monitor-2025
Levytska, I. (2026, 3 de julho). Russia begins mass deployment of autonomous AI-equipped drones in Zaporizhzhia. Ukrainska Pravda. https://www.pravda.com.ua/eng/news/2026/07/03/8042223/
Marcon, J. (2025, 5 de novembro). Autonomous weapons systems and their threat to security in Latin America. Stop Killer Robots. https://stopkillerrobots.medium.com/autonomous-weapons-systems-and-their-threat-to-security-in-latin-america-3f0f38cd6aa3
Sehlac, Latin America and the Caribbean Human Security Network, Fundación para la Paz y la Democracia, Technology and Community, & Centro de Estudios Política Internacional. (2024). The dangers of autonomous weapons systems from a Latin American perspective. https://docs-library.unoda.org/General_Assembly_First_Committee_-Seventy-Ninth_session_(2024)/78-241-SEHLAC-EN.pdf
Stop Killer Robots. (n.d.). Stop Killer Robots: Campaign for a treaty on autonomous weapons. https://www.stopkillerrobots.org/
Varella, L. (2026a, 6 de setembro). Editorial: Ready for negotiations. CCW Report, 14(4), 1–2. Reaching Critical Will. https://reachingcriticalwill.org/disarmament-fora/ccw/2026/laws/reports/18216-ccw-report-vol-14-no-4
Varella, L. (2026b, 6 de setembro). Report of the meeting of the Group of Governmental Experts on Lethal Autonomous Weapons Systems. CCW Report, 14(4), 3–23. https://reachingcriticalwill.org/disarmament-fora/ccw/2026/laws/reports/18216-ccw-report-vol-14-no-4
Comments