University of Maryland's Defense Department research on modeling "population complexity" with OSINT and LLMs for decision support — a federal contract (USAspending)

This contract follows a familiar pattern: a government body commissioning research from a university. The awarding office, the Defense Department's Washington Headquarters Services, supports the operation and administration of the DoD headquarters, and one of its components has turned to a university to bring in outside expertise. The recipient, the University of Maryland, College Park, is a public research university close to the nation's capital and a natural home for this kind of government-linked research.

At the center of the work are OSINT (open-source intelligence—drawing insight from sources anyone can access, such as news, public records, and web content) and PAI (publicly available information). The stated purpose in the original data is to combine these with AI tools and large language models (LLMs—AI trained on large bodies of text that can generate and analyze natural language) to model "population complexity" in support of decision-making. It is one example of the broader effort to apply LLMs, on a research basis, to government analytical work.

What this contract illustrates is that the government is carefully examining—at the research stage—where and how rapidly spreading generative AI and large language models might be used. Framing it as a research contract rather than an operational deployment suggests an intent to gauge both the usefulness and the limits of these tools. Because the theme involves analyzing public information and modeling people, questions about privacy and the handling of information can arise as a general matter; however, the specific scope, methods, and deliverables of this contract are not described in the original data, so no conclusions can be drawn here.

Viewed more broadly, this sits within the wider trend of AI procurement and AI research across the federal government. A single contract published on USAspending lets anyone trace which technologies the government is testing, with which universities, and at what scale—offering citizens a window into both how tax dollars are spent and the direction of technology policy.