ABSTRACT
Life-altering impacts are sometimes the outputs from artificial intelligence (AI) and automated decision systems (ADS) used by governments and organizations seeking to improve public services. In contrast to the design, development, or adoption of AI, we establish the procurement of high-risk AI and ADS as an unexplored greenfield that needs discerning attention and innovation. We emphasize that this attention has to originate from an unrealized fiduciary duty to the public. In this chapter, we highlight the relevant AI adoption landscape, the problem of automation without citizen remedy, and the pressures of AI infallibility as a contextual preface to selected challenges for the procurement of these systems in high-risk domains. Transdisciplinary collaboration teams and sandbox testing challenges are then described. In the teams and sandbox sections, the rationale for government agencies and standards developers to work together in a purposeful manner to innovate procurement processes is explained.
Brief summaries follow of the published AI governance standards and the ethical AI certification program offered by the Institute for Electrical and Electronic Engineers (IEEE). Next, the new category of international consensus-based standard is introduced: the IEEE P3119 Standard for the Procurement of AI and Automated Decision Systems. The standard’s primary aim is to give government agencies, AI vendors, and organizations the opportunity to strengthen their requirements for AI procurement projects serving the public interest. Adding to existing local, national, and transnational regulatory regimes, the new standard will offer complementary normative process guidance that can be integrated across various stages of acquisition from pre-procurement (solicitation) to contract monitoring. An overview of the draft process model, the guidance offered in its five processes, and the tools are provided. Finally, its alignment to risk management practices and the founding of the AI Procurement Lab are reviewed. We conclude the chapter with an acknowledgement to the demanding work by volunteers and partners helping to raise the global standard in responsible AI procurement.
