In This Story
The Department of Defense (DoD) faces unique data licensing challenges when acquiring artificial intelligence (AI) solutions from the private sector. This paper provides an alternative framework to the traditional data licensing strategies to better address the unique challenges of acquiring AI solutions.
In a traditional procurement, DoD identifies data and software to be delivered under the effort, which DoD will fund (at least partially), and thus will expect to obtain a license in that data and software. In this traditional context, the ultimate objective of negotiation is to avoid so-called “vendor lock” in the procurement and sustainment of the developed solution. The data (including software documentation) and software deliverables are typically designed to allow for the government’s use in subsequent competitions for further development, production, and sustainment, or, in some situations, for organic testing and modification. The data deliverables are thus meant to provide the information necessary to produce an equivalent part, system, or application.
AI-based solutions present a challenge to this paradigm, however. Cutting-edge AI technologies are largely developed in the commercial sector at private. To achieve the Responsible AI principles, the government customer must understand how a technology works and how the results of the technology are derived, not simply how to produce more units. As a result, the types of data required to assure that delivered AI-enabled technologies are responsible, equitable, traceable, reliable, and governable2 go beyond the delivery of end-solutions. This presents a challenge for companies attempting to enter the Defense market through commercial business models, standard seat or enterprise licenses, or “As–a–Service models.” Providing source code, proprietary datasets, software architecture, or background IP to any entity, let alone the government, is far outside of the norm for these commercial AI firms.
Thus, to attract and integrate cutting-edge AI solutions, while adhering to Responsible AI principles, DoD—specifically DoD acquisition professionals—should implement a framework for assessing license needs, and craft custom use licenses that balance the need for access and use of data and software for RAI purposes against industry partners’ desire to protect their IP and maintain standardized commercial licensing practices.