ABSTRACT
AI technologies will continue to expand in all domains, including in military applications whose criticality is high. The level of exigence for governance and qualification of systems is particularly high in this domain, while the field of AI is well-known for its burstiness, enthusiasm, and fast pace. Currently, deployed AI systems sometimes present a lack of quality, legal issues, and undocumented accountability frameworks among other defects. This lack of awareness is sometimes identified as a “no-go” in the defense domain. In this chapter, an existing EU-supported AI application assessment framework, ALTAI (HLEG, 2019), is promoted by reviewing three military use cases and highlighting their relevance and shortcomings. It is claimed that ethics assessments do bring an added value to AI development and that potential solutions such as “explainable AI” or “exhaustive tests”, even if desirable, are neither sufficient nor necessary to decide to use AI systems.
