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

Dialogue with allies can overcome differences over the ethical use of AI in defense – discussion moves principles into practical implementation.

Stanley-Lockman, 2021 - Center for Security and Emerging Technology [Zoe CSET Issue Brief August “Responsible and Ethical Military AI Allies and Allied Perspectives” https://cset.georgetown.edu/publication/responsible-and-ethical-military-ai/ Acc. 4/21/22 TA]

Additionally, differences between allies’ views on responsible and ethical AI in defense may also stem from the extent to which other countries apply civilian AI policy and regulation frameworks to their own defense approaches. Although overviews of government principles and policies for ethical civilian AI are not discussed at length here, they become more visible in other countries, as well as the EU. Some of these aspects are encouraging—for example, German industry’s voluntary compliance with trustworthy AI principles, to the extent they overlap with defense, due in part to the fact that there is no equivalent defense framework they can follow and implement themselves. Other entry points of civilian concepts into the defense realm include the Australian concept of contestability, Canadian DND subjectivity to the algorithmic impact assessment, and the choice of some EU countries to apply the General Data Protection Regulation to their defense sectors. These are both related to accountability as well as privacy, which are key differences that should be tracked even though allies overwhelmingly agree on the importance of ethics and safety for AI in defense. Still, similarities between defense stakeholders include the view that militaries could not only inject new risks into their operating environments, but also expose their own organizations to risk if they leave ethical and legal questions unaddressed. Countries may have different ways to define and measure these ethical risks, as Table 1 implies, and as is detailed in the appendices. Overall, though, there is an implicitly common approach which recognizes that they must contend with the associated technical, legal, political, and moral risks from the front end of AI development. More concretely, they also agree that the way to implement responsible AI involves technical measures tied to safety and security, as well as procedures that make the legal context by which they abide as clear as possible. Overall, pathways for leveraging shared views to advance the implementation of responsible AI should include learning from allies whose views emphasize both human responsibility and responsible state behavior extending beyond minimum legal obligations. Addressing these issues is not only a question of good engineering practices, but is also an exercise of responsible state behavior.188 In a narrow sense, responsible AI can refer to ensuring that AI systems enter into human-centric frameworks that are defined by humans to maintain human agency and responsibility. More broadly, though, it is notable that some allies see preserving freedom of action as part of a vision of responsible AI that encompasses responsible state behavior. Having a legitimate basis for military action is a feature of responsible state behavior, with civilian government oversight of militaries at its core to maintain accountability at home and abroad. This enters into the language of responsibility because operating in coalitions under multinational mandates can also confer international political legitimacy to operations.189

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