U.S. Army contract to ColdQuanta to protect AI positioning, navigation, and timing systems from adversarial attacks — a federal contract (USAspending)

Positioning, navigation, and timing (PNT) is the foundational technology for knowing where you are, which way you are heading, and exactly what time it is, with GPS as its best-known example. Much of modern life depends on this accurate sense of place and time, from military operations and logistics to communication networks and the time synchronization behind financial transactions. In recent years, AI and machine learning have increasingly been combined with PNT to keep accuracy high even when signals are degraded or jammed. Yet relying on AI can itself introduce new weak points.

This contract is aimed at guarding against exactly that weakness. AI faces a threat known as an adversarial attack, which uses input data crafted to be hard for humans to notice in order to trick the AI into making wrong decisions. In a field like PNT, where a small error can lead to large consequences, an AI fooled into misreading position or time could have serious effects. The research seeks to defend AI- and machine-learning-based PNT systems against such attacks and to strengthen their resilience so that they keep functioning even when targeted. It sits squarely within the field of secure AI, meaning AI built to withstand attacks.

The contract is meaningful because it turns a broadly important theme, balancing the usefulness of AI with its safety, into a concrete procurement. As AI is embedded more deeply into society's core systems, protecting the AI itself from attack grows more important. By focusing on PNT, a piece of highly public infrastructure, this case offers a window beyond the defense sector into how public funds support research on using AI safely. The specific results and methods of the research are not described in the source data.