AI Use by the US Government Is Ballooning. And the Lack of Transparency Is Troubling | Nathan E Sanders and Bruce Schneier

Why It Matters

The rapid expansion of government AI without robust oversight threatens civil liberties and public trust, while also missing an opportunity to harness AI for genuine public benefit.

Key Takeaways

• 3,611 AI projects across U.S. agencies, up 70% YoY
• AI slated for life‑or‑death decisions in health and nuclear sectors
• Disclosures provide only brief descriptions, no impact assessments
• Public input is limited; only DOJ case invites comment
• France and Canada offer more transparent, consultative AI frameworks

Pulse Analysis

The Office of Management and Budget’s latest AI inventory underscores a seismic shift in how the U.S. government leverages machine learning. With 3,611 active or planned projects—a 70 percent increase from the prior year—federal agencies are embedding algorithms into everything from grant reviews to prison classification. This acceleration reflects both the allure of cost savings and the pressure to modernize legacy processes, yet the brief, one‑sentence entries offer little insight into data sources, model validation, or potential bias. For stakeholders, the sheer scale raises red flags about oversight capacity and the risk of automated decisions affecting fundamental rights.

High‑stakes applications illustrate the tension between promise and peril. The Department of Veterans Affairs plans an AI‑driven suicide‑risk assessment that listens to crisis‑line calls, while the Department of Energy tests autonomous control of nuclear reactors. Simultaneously, the Federal Bureau of Prisons is developing a system to pre‑emptively label new inmates as high‑risk. These use cases could improve outcomes if rigorously tested, but the inventory’s lack of risk‑assessment documentation and public comment mechanisms fuels distrust. Transparency gaps make it difficult for watchdogs, legislators, or affected citizens to evaluate whether safeguards are adequate or whether the technology reinforces existing inequities.

International examples provide a roadmap for responsible deployment. France’s 2016 Digital Republic Act mandates algorithmic transparency, human‑review rights, and mandatory notifications for automated decisions. Canada’s 2025 AI registry couples public disclosure with a federal directive requiring detailed risk‑scoring and stakeholder consultation. Adopting similar standards—mandatory impact assessments, clear appeal pathways, and structured public comment periods—could restore confidence while allowing the government to reap AI’s efficiency gains. Without such reforms, the rapid proliferation of AI in federal operations may outpace the very democratic safeguards it aims to support.