Amid the ongoing public dispute between the Department of Defense and Anthropic over government powers regarding technology use for mass surveillance, another U.S. agency is proactively working to prevent future conflicts. How are they doing this? By revising government procurement rules.
Procurement—the process governments use to acquire goods and services—has long been a strategic means to advance policy objectives. By determining where and how to allocate funds, the government effectively conveys its political and social priorities. Therefore, it’s imperative that government spending supports initiatives that benefit the public, such as open-source software development, interoperability, and the right to repair. Conversely, funding should be withheld from dubious contractors with insufficient security practices.
The General Services Administration (GSA), responsible for federal procurement, has proposed new rules aimed at furthering one main policy goal: the support of “ideologically neutral” American AI innovation. However, these guidelines extend far beyond that single ambition.
In comments submitted alongside our partners at the Center for Democracy and Technology, the Protect Democracy Project, and the Electronic Privacy Information Center, we highlighted significant concerns about the GSA’s proposed guidelines. These regulations include broad provisions that could compromise the safety and utility of AI tools. If adopted, they would become standard elements of all federal contracts. You can read the full comments here.
One of the most troubling stipulations is that contractors and service providers must license their AI systems to the government for “all lawful purposes.” Considering the government’s broad interpretations of legal standards, its capacity to exploit loopholes for surveillance, and its history of engaging in unlawful spying, we urgently need robust legal safeguards to prevent the government from amassing personal data and potentially using routine bureaucratic information for punitive measures.
Additionally, the draft rules compel that “AI System(s) must not refuse to produce data outputs or conduct analyses based on the Contractor’s or Service Provider’s discretionary policies.” This means if a company’s safety measures could restrict responses to a government inquiry, those safeguards must be disabled. Considering the widespread public apprehension regarding AI safety, it seems unwise to limit the protections a company considers necessary.
There are numerous other issues within the draft rules, including overly ambiguous “anti-Woke” clauses. However, the central concern stands clear: much of the proposed guidelines fail to align with the public interest in utilizing American tax dollars to endorse privacy, safety, and responsible tech development. The GSA ought to reconsider its approach and start anew.
