Efforts to reduce risk in the release of artificial intelligence enabled systems

Abstract

There has been great attention paid to the risks with the proliferation of Artificial Intelligence across both commercial and defense sectors. While the National Security Commission on Artificial Intelligence report highlighted the need to “establish justified confidence in AI systems,” it did acknowledge that “the United States will field safe and reliable AI…” through the execution of existing procedures. This acknowledgment reinforces the prevailing thought from Combat Capability Development Command Armaments Center that there are robust processes in place, such as the Materiel Release (MR) process. The authors spent years engaged with stakeholders, board members, coordinator, and chair to collect the prevailing concerns, distill them into risks, and identify mitigations that would enable the board to properly and thoroughly review an AI enabled system. This paper will outline the findings of this analysis, where the MR Process may be altered or tailored to account for AI, and suggestions for stakeholders to reduce risk. The culmination of this work is a roadmap to ensure not just responsible, but safe, suitable, and supportable systems continue to be fielded by the US Army.

Ähnliche Arbeiten