CIA Deploys AI to Produce Open-Source Intelligence Briefs for Local Police
The CIA is using generative AI, specifically GPT-4.1, to create open-source intelligence reports that are sent to Washington State’s fusion center for distribution to local law-enforcement agencies.
According to records secured by Reason, the CIA’s Open Source Enterprise (OSE) has begun producing open-source intelligence analyses with the aid of generative AI, naming GPT-4.1 as the model employed. OSE staff input articles from major foreign outlets, run keyword relevance algorithms, and use the AI to translate titles and summarize the pieces, with analysts reviewing each step for fidelity. During the current World Cup, the unit generated daily briefs on security conditions in Canada and Mexico, which were transmitted to the Washington State Fusion Center for onward distribution to local police forces.
The fusion center, part of a nationwide network created after September 11, has a mixed reputation for efficiency and accuracy, and it redacted some of the CIA documents for security reasons. This development illustrates a broader trend of U.S. security agencies integrating AI tools—from ICE’s facial-recognition systems to the FBI’s watch-list software requests—into everyday operations, effectively expanding the paperwork pipeline to local law-enforcement partners.
Why it matters
It shows how AI is being used to automate intelligence sharing between federal agencies and local police, raising questions about oversight and data accuracy.
In this story
