Procuring Algorithmic Transparency
UVA Law Faculty Affiliations
Artificial intelligence tools (“AI”) have great potential to improve government functions and efficiency. These algorithmic tools, protected by trade secrecy, are often developed by private vendors and sold to government agencies. As a result, there is a growing symbiotic relationship between private sector sellers-of-technology and government purchasers-of-technology in almost every governmental organization, including at every stage of the criminal justice system. However, as the government increasingly relies on private vendors to supply its technologies and the attendant secret algorithms that aid decision-making, the public’s call for transparency presents significant challenges. That is because private vendors’ assertion of trade secret rights in these technologies seemingly conflict with the public’s need for disclosure.
Using the criminal justice system as an illustration, this Article conducts an in-depth analysis and posits that the existing theoretical framework of trade secret law presents both the problem and the solution to the transparency problem. Our analysis of the problem is the first to focus on the point at which a government agency initially decides to acquire its software. We suggest that many of the relevant concerns regarding excessive secrecy and essential government transparency could be resolved through negotiation of appropriate disclosure terms and conditions that are supported not only by trade secret law, but by government procurement and contracting law and policy. Accordingly, the Article offers a novel transaction-by-transaction procurement-based proposal based on a negotiated approach that carefully weaves criminal law and procedure, constitutional law, contract law, intellectual property law, and procurement law.