Free beta access. 50 free AI tokens each week. Optional top-ups available. See what is included
Opening Word document…

The EU has the capacity and motivation to get ahead on AI regulation --- EU regulation is critical to ensuring fail-safe mechanisms that prevent unpredictable AI

Boulanin et al ’20 – Vincent, Senior Researcher leading SIPRI’s research on emerging military and security technologies., Netta Goussac, Associate Senior Researcher within the Armament and Disarmament research area with particular expertise in legal frameworks related to the development, acquisition and transfer of weapons., Laura Bruun, Research Assistant working on emerging military and security technologies., Luke Richards, Research Assistant working on emerging military and security technologies.,(“ RESPONSIBLE MILITARY USE OF ARTIFICIAL INTELLIGENCE Can the European Union Lead the Way in Developing Best Practice?”, 11/20, pg.13-16, https://www.sipri.org/sites/default/files/2020-11/responsible_military_use_of_artificial_intelligence.pdf) – sel

Ethical acceptability: Defining shared principles for human-centric artificial intelligence in the military

Human agency as a central issue in responsible military use of artificial intelligence

AI raises many different ethical issues that are relevant in both the civilian and military sectors. These issues range from how data serving the development and training of AI systems is sourced, used and protected, to the biases and assumptions that appear in AI programming, to the potentially dehumanizing effect of reducing the real world to data points, to the consequences of non-transparent or inexplicable decision making for accountability and responsibility.50 Many of these issues feed into a higher-level discussion on how AI is challenging human agency. In this regard, the EU member states seem to already agree on the fundamental premise that human agency—be it in the civilian or military sector—is necessary for responsible use of AI. Finland is one EU member state that has given this issue a lot of thought. In its submission to the GGE on LAWS 2020 it stated, for example, that ‘the easier an advanced technology is to apply, the easier it will be to use it for harmful purposes. In the future, we might well see armed autonomous civilian capabilities used for military purposes’, arguing for the necessity of establishing a framework for human involvement.51 AI should remain a tool for humans to make decisions, and it should not strip humans of their obligation and ability to do so. That principle is a cornerstone of the strategy of the European Commission on ‘human-centric AI’.52 Members of the European Parliament also agreed in 2018 that the litmus test for determining if an autonomous weapon would be acceptable is whether it remains under ‘meaningful human control’.53 In the context of the GGE’s debate on LAWS, EU members states also supported the adoption of the guiding principle which states that ‘human responsibility for the decision on the use of force must be retained’ (see box 2.1).54 Many of the EU states have made the concept of human control a pivotal point of their national position in the debate on LAWS, such as Germany that has stated for instance that ‘no weapons system may, on the strength of its algorithms, entail the risk of overriding a human de-activation command’ (see figure 3.3).55 While EU members agree that meaningful human control should be maintained throughout the entire life cycle of the weapon, the question of how human agency needs to be exercised in practice, however, remains unresolved.56 While some proposals have been made in the context of the discussion of GGE on LAWS, they are limited in scope as they do not have the intent to guide other possible military uses of AI, from logistic and maintenance to mission planning, training and recruitment of personnel.57 The general principles and recommendations laid out by the AI HLEG could, in this regard, provide a useful baseline to think about how human-centred AI could look like in the military sector, beyond the sole case of LAWS. EU member states would have a lot to gain from engaging in an intra-European expert discussion on human control. First, on the issue of legal compliance for example, it would allow them to discuss, deepen, align or, at least, remove the potential for conflict in their reflections on the case of LAWS. Second, it would allow them to find a clear direction and to determine their limits in developing collaborative defence acquisition projects. Fostering intra-European discussion on human-centred artificial intelligence in the military Deliberations on human-centred AI in the military could take place in different forums at the same time. The choice of forum should be guided by concrete political aims. The Council of the EU already has a number of preparatory bodies that would allow delegates from member states to coordinate their views on human control in relation to discussion on LAWS at the UN in Geneva, namely, CONOP, CODUN and EUMCWG. The regular meeting of these bodies provides an opportunity for the EU External Action Service to identify the current common position of EU member states on this issue. The meetings, however, are meant to cover a wide range of topics every time, which limit the possibility for states to discuss topics at a very granular level. It could be meaningful in that regard to organize a series of extra meetings specifically dedicated to the issue of human control. When it comes to developing more general political guidelines or technical standards for the development of joint defence acquisition or capability development projects involving AI, then the EDA and the European Commission could both play a role. EU member states could task the EDA to coordinate an expert reflection on human-centred AI for defence, which would require technical, military and arms control experts. Such an initiative would need to connect with both the work on human-centred AI in the civilian sector and with the discussion on human control within arms control forums. The European Commission could utilize the EDF to support multidisciplinary research projects that would explore issues around human control at a more technical level and feed into the development of AI systems that can be deployed responsibly It would also be valuable if EU member states could engage in the issue at a higher political level and grapple with more fundamental ethical and political issues. Such a discussion could take place in the context of an ad hoc expert group of the European Parliament and directly at the level of the ministers of foreign affairs and defence and the EU High Representative of the Union for Foreign Affairs and Security Policy at the European Council.

Sign in to recut, update, or save this card.