US Intelligence Quietly Launched AI to Pinpoint Where Almost Any Video Was Filmed in Under 5 Minutes

A U.S. intelligence program called LocUS, developed by the Intelligence Advanced Research Projects Activity (IARPA), aims to use artificial intelligence to determine the location of videos, images, or audio clips without relying on embedded metadata. The system is designed to assemble fragments like traffic sounds, electrical outlet shapes, and room acoustics to pinpoint a location within 820 feet in less than five minutes. The program was quietly released nearly five months ago and is intended to help intelligence and law enforcement locate hostages, identify human trafficking sites, or trace online videos. It also highlights the difficulty of making multimedia truly anonymous.

• LocUS is an intelligence-grade system that combines visual and audio information, including environmental sounds, to determine locations.
• The initial target is for 90% of predictions to fall within 820 feet of the true location.
• The system must work indoors and outdoors, across vast search areas, and process degraded material like blurry video or low-quality images.

The program requires systems to fuse all available information into a single location prediction and excludes biometric methods like facial or voice recognition. IARPA plans to test submitted technologies using previously unseen multimedia held in sequestered datasets. The software must operate autonomously, without human guidance, and is designed to stress-test weaknesses in geolocation models, particularly in unfamiliar locations like indoor environments.

Public responses show researchers were already organizing around the problem, with some highlighting the importance of datasets like those from TraffickCam, which has been downloaded by over 250,000 users. The program is expected to be tested over 15 months, with five independent events, and is designed to improve geolocation capabilities beyond imagery-only methods. However, the final performance or deployment of LocUS may remain undisclosed, making it an active intelligence capability without full public knowledge of its progress.

AI-generated summary of the original article — see the source for the full story.