Symposium on the CCW GGE LAWS: The Human Element in the GGE LAWS 2024-2026 Report

[Anna Andersson is a Senior Researcher in International Law at The Swedish Defence Research Agency (FOI). Anna was a member of the Swedish delegation to the GGE LAWS 2023-2026. This post is, however, written in her capacity as a researcher and does not reflect Sweden’s position or understanding. This post is the fourth in a series of analyses on aspects of the recent adoption of the 2024-2026 CCW GGE LAWS report].

Key reasons behind the international discussion on lethal autonomous weapons systems (LAWS) are concerns related to what LAWS may mean for human control over the use of force, human dignity in armed conflict, and for compliance with international law. A central issue has therefore been to which extent existing international law requires human control, and whether this should be explicitly required by new legally binding rules. This was a persistent issue which divided States throughout the work of the Group of Governmental Experts on Lethal Autonomous Weapons Systems (GGE LAWS) within the Convention on Certain Conventional Weapons (CCW). In fact, the so-called “human element” was the last element to be resolved in the GGE’s September session, just before the adoption of the 2024-2026 report as a whole, which contains a set of elements for an instrument for LAWS. 

Paragraph 34 of the report sets out that to uphold compliance with international law applicable in armed conflict, control and human judgment is needed, and mentions factors to be taken into consideration in this exercise. Thereafter, paragraph 35 follows with a non-exhaustive list of relevant measures. In this post, I will seek to analyse the development of the discussions on the human element, through a focus on the three main notions of human involvement. Thereafter I will analyse paragraphs 34 and 35 in light of the rolling texts and the September discussions, to highlight what is and what is no longer included in the text.

The development of human control

There were predominantly two opposing approaches to formulations of human involvement. On the one hand, those who wished to have an emphasis on human control. On the other, those who wished to separate control from the human being. This division, and attempts to resolve it, characterized the discussions throughout. Three major shifts can be identified in the discussions. They follow how broad support formed around three main notions at different stages of the discussions, from ‘meaningful human control’, via ‘human judgement and control’, to ‘control and human judgement’. In the end, the Group could only find consensus on the latter.

The power of a notion: meaningful human control

Already before the GGE LAWS was established (in 2016), the need to retain ‘human control’ or ‘meaningful human control’ of LAWS was recognized by many States and observer delegations in the CCW (see for example a proposal ahead of the 2013 State Party meeting and the 2014 report of the informal meeting on LAWS). ‘Meaningful human control’ had been coined by the civil society organisation Article 36, as a policy term providing that human control is needed and must meet a certain threshold.

The formation around ‘meaningful human control’ presents the first shift in the discussion of the human element. It gained support by a large group of States from all parts of the world (see for example here, here and here), as well as expert- and civil society organizations (including in the Ban (later Stop) Killer Robot Campaign), and long remained the predominant notion in the GGE. This notion has been described as “intuitively appealing”, partly, because of its accessible language and that it places the human being at the centre of an issue that challenges her role in warfare, and partly because it is flexible enough to gather actors with a broad spectrum of views on what it actually entails (UNIDIR and Mauri, 2022). 

Several States, the ICRC and civil society organisations, including Article 36, that expressed support for ‘meaningful human control’ remained open for other terms than ‘meaningful’ that could delineate the required or desirable threshold, such as ‘appropriate’ or ‘necessary’. In addition, several other States, supported ‘human control’ but were agnostic to ‘meaningful’ as the qualifier. Whereas Russia rejected ‘meaningful’ as inappropriate but used ‘human control’. Accordingly, in 2018, ‘human control’ was considered as the core notion for advancing the discussion.

While not straight out rejected, notions of human control met continuous opposition from States that had concerns about developing this as a new legal concept. Among them, India and the United States, preferred the term ‘human judgment’.

This divergence persisted and impacted which formulations could be agreed on for the Group’s reports. In the 2019 report the Group concluded that “Human judgement is essential in order to ensure that the potential use of … [LAWS] is in compliance with international law, and in particular IHL”. However, it was not included in the Guiding Principles, adopted as an annex to that report. In the 2023 report States agreed that “Control with regard to … LAWS is needed to uphold compliance with international law, in particular IHL, including the principles and requirements of distinction, proportionality and precautions in attack.” Accordingly, the Group explicitly linked both human judgement and control to international humanitarian law (IHL) compliance, but separately and in different years and reports.

Seeking a compromise: human judgement and control

When work on the Group’s mandate to formulate a set of elements for an instrument for LAWS began in 2024, the Chair built on previously used language, as the GGE has done frequently, but with a view to advancing it and seek a compromise. The first rolling text (of 26 July 2024)  introduced the notion of ‘appropriate control and human judgement’. For the first, a text that the Group as a whole focused its attention on combined these two concepts into one notion. It was not without resistance, but with sufficient support for development. The next version of the rolling text (of 8 November 2024), maintained this combined notion. As focus centred around the rolling text, ‘meaningful human control’ began to lose traction as the main notion. But it was the coming version of the rolling text that brought about the next major shift in the discussion.

In the rolling text of 12 May 2025, the notion on human involvement was changed to ‘context-appropriate human judgement and control’. Rather quickly, a vast support materialized behind this new notion. This movement of support presents the second significant shift in the discussion on human involvement.

Many States held that they continued to prefer ‘meaningful human control’ but expressed that they could support or accept ‘human judgement and control’. Others preferred the new notion. A number of states offered a joint understanding, and described it as reflecting human involvement more broadly. Motivations of support included that it captured the same essence, that its formulation was more suitable for an instrument, and that it could help bridge divisions and find a way forward. For some States it constituted a big leap to change from ‘meaningful human control’ to ‘context-appropriate human judgement and control’. It was a step that was taken in a spirit of compromise to meet the strong opposition. 

A relevant aspect was that ‘human judgement and control’ could be read both as ‘human control’ and as simply ‘control’, distinct from human. For some it provided a solution to the division in the Group, but for others this constituted a serious concern. A relevant aspect was that some delegations had suggested that control could be exercised through software. At least one state also voiced concern that combining control and judgment could cause conflation of distinct concepts. This was subject to intense discussions in 2025 and the March 2026 session.

From the 12 May 2025 version to the last rolling text, of 5 June 2026, ‘human judgment and control’ was maintained, and over time the support grew. In the March 2026 session, it appeared that almost all States supported or could accept it. However, the United States rejected this notion and it appeared consensus could thus not be reached on this. It is against this background that the third shift should be seen. 

The final stretch: control and human judgement

The third shift occurred when consensus was formed around ‘control and human judgement’ in the last session under the GGE’s 2024-2026 mandate. The shift started during the informal hybrid summer sessions, and ahead of the September session, with the Chair presenting this notion in his Additional Suggestions. Time was of the essence and influenced States’ willingness to make further compromises.

This notion might not have had many supporters in and of itself, but it was one that all States could live with. It was a difficult compromise for many States. It entailed another big leap for States who had reluctantly moved from ‘meaningful human control’ or ‘human control’ to ‘human judgement and control’. ‘Control’ was now separated from ‘human’. The compromise for the United States was the combination of ‘control’ and ‘human judgment’, that it had previously rejected. These compromises indicate a genuine willingness for the Group to succeed.

To uphold compliance with international law, control and human judgment is needed. The first sentence of paragraph 34 thus reads:

34. To uphold compliance with international law applicable in armed conflict, in particular the principles and requirements of distinction, proportionality, and precautions in attack, control and human judgement with regard to LAWS are needed. …

This essentially brings together what was stated on human judgment in the 2019 report and on control in the 2023 report in relation to compliance with international law, in particular IHL, to one notion. It clarifies that both control and human judgment are needed to uphold compliance with international law.

In contrast to the mentioned reports, this sentence uses ‘international law applicable in armed conflict’, which derives from the preamble of the CCW. This was a late change on the evening of the last day of the September 2026 session from the reference to ‘international law, in particular IHL’ used in the rolling texts. It of course includes IHL, but also other applicable areas of international law. In contrast, the third sentence of paragraph 34 uses only ‘applicable IHL’. Both terms appear in other elements of the report, and it is not clear whether the difference was always intentional.

After the first sentence on control and human judgment is set out, paragraph 34 goes on to state:

… Human beings exercise such control directly or indirectly, including through measures taken before and/or during the use of LAWS. Regardless of how control is exercised, human beings need to exercise judgement to determine those measures necessary to ensure compliance with applicable IHL, including the principles and requirements of distinction, proportionality, and precautions in attack, and to implement such measures in practice. …

The second and third sentences of paragraph 34 was important for several States to agree to placing ‘control’ before ‘human’ in the first sentence. It means that this notion does not stand on its own, paragraph 34 and paragraph 35 should be read in full and together and was described as a package deal in the final discussions. Notably, the second sentence set out that humans can exercise control directly or indirectly, possibly through software or measures listed in paragraph 35. The ‘and/or’ was particularly hard to agree on, as at least one State wanted to remove the ‘and/’, while that temporal aspect was key for other States.

Discussions lingered over whether it should be ‘to determine what measures necessary’ (as reflected in the 3 September draft report), ‘to determine the measures necessary’,  or the compromise ‘to determine those measures necessary’, which gained consensus at the last minute, with India pronouncing its national understanding of it. Whereas these discussions partly reflected substantial concerns, and partly appeared to be related to language issues, there were also indications of obstruction by spending time on certain issues at the cost of others.

Paragraph 34 is the only element in which ‘control and human judgement’ appear in the set of elements. This stands in contrast to the rolling texts which made more frequent use of ‘human judgement and control’ (and similar notions). In all rolling texts except the last, ‘human judgement and control’ formed part of three to five elements, including in an element with a prohibition on employing LAWS that operate without it. The human being was thus significantly more central to the GGE’s discussions than the 2024-2026 report reveals. 

Considerations of context

The last sentence of paragraph 34 refers back to ‘control and human judgement’ and set out that:

… In doing this, a range of factors need to be considered, including the operational context and the characteristics and capabilities of the weapon system. 

This part may be understood both as a qualifier and as a mechanism for flexibility. It reflects that IHL obligations may depend on the domain (as land-, naval- and aerial-warfare have certain respective specific regulations), that its requirements depend on the area of operations and the situation at hand (such as the presence of protected persons or objects), and the specific characteristics and capabilities of the weapons available and used. 

The need to consider context in the exercise of control has been recognized from the beginning of the GGE’s discussions, and has not been controversial. Yet, time and attention at the end of the discussions were spent to reformulate and move this part around. In all but the last of the Chair’s rolling text versions considerations of such factors were formulated simply as ‘appropriate’ or ‘context-appropriate’, and was generally supported as such. It was late in the March 2026 session that some delegations suggested to unpack and reformulate this in order to seek compromises on the deadlock around the notion of ‘context-appropriate human judgement and control’. This led to the return of the formulation of the 2019 Guiding Principle (c) on human-machine interaction, which clarifies that the weapons characteristics as well the operational context is to be considered. 

Relevant measures

The second element dealt with in this post is paragraph 35 of the GGE’s report. This reads as follow:

35. Relevant measures include, inter alia: 

(a) That LAWS are operated under a responsible chain of human command and control; 

(b) Limiting the scale of the operation, types of targets, duration, and geographical scope, as appropriate; 

(c) Measures to intervene in the operation of LAWS, to the extent possible or feasible, including to deactivate or neutralize LAWS in a timely manner.

The chapeau provides that what follows is a non-exhaustive list of examples of relevant measures. Its opening words, ‘Relevant measures’, were a compromise to refer back to paragraph 34. The list can thus be read as outlining measures that contribute to operationalise control and human judgement.

The list in paragraph 35, especially (a) and (c), together with its clear link to paragraph 34 and IHL compliance, provide a step forward from the Guiding Principles and the 2023 report. The Guiding Principles note that human-machine interaction may take many forms, but did not mention limitations of LAWS or practical measures of control. The 2023 report mentions the limitations in (b) but without the linkage to human agency and IHL requirements that paragraph 34 provides. 

However, the list falls short of what could be expected from the last three years’ discussion on measures and limitations. The number of measures that gathered consensus are fewer than the measures that were discussed in the GGE. A comparison of the Chair’s rolling texts from 2024 to 2026 (see for example versions 8 November 2024,  18 December 2025 and 5 June 2026), the draft report of 3 September 2026 and the Chair’s Additional Suggestions demonstrates that a number of measures were excluded along the way, some of which at the very end of the last session. This was partly due to the time constraint, partly due to elements of obstruction.

Since paragraph 35 is non-exhaustive it is worth mentioning these measures.  Measures that were excluded in earlier phases include: limit the number of engagements that LAWS can undertake; limit the operation of LAWS to a defined perimeter in particular to reduce the likelihood of harm to civilians or damage to civilian objects; restrict the use of LAWS against objects that are military objectives by nature; and to ensure appropriate training and instructions for human operators of LAWS.

Two measures, containing three aspects, were included in the last rolling text, the draft report and the Chair’s Additional Suggestions, but were excluded during the GGE’s last session. These were a part in (a) on assessment of legal obligations by a human, and of taking ethical considerations into account, and former (c) providing that LAWS’ target and mission parameters cannot be substantially modified unless such modification is executed or authorized by a human. Given how long these measures were part of the discussions, and their importance for some States, they may surface again in the future.

Conclusion 

From the beginning, the formation around specific notions on human involvement came to have a defining impact on the discussions and its results, and sometimes the discussions on the notions overshadowed the issues they were meant to advance. The Group have indeed discussed how human control relates to rules and principles of IHL and how it may be implemented in practice. But it has also spent much time and effort focusing on the notions as such.

IHL presupposes that it is human beings who plan or decides upon an attack, and regulates this. The legal literature demonstrates an ongoing discussion on to which extent those persons may use LAWS in these tasks (see e.g. Ekelhof, 2018, Jensen, 2020, MacFarland, 2022, Sehrawat, 2025). Paragraph 34 bring some guidance to this discussion. At the same time, it does not elaborate in detail on what is required of the respective applicable principles and requirements in terms of human involvement in relation to LAWS. This may be reflective of the nature of diplomatic discussions as well as of the complexity of the issue at hand. In any case, issues remain.

It follows that further work is needed to deepen the understanding of how the notion eventually agreed on should be understood in relation to specific rules and principles. It must not be seen in isolation, or only in light of the rest of paragraph 34 and 35 as dealt with in this post, but as an reaffirmation of the centrality of human beings to compliance with international law, in particular IHL, and in light of all elements included in the report.