The CIA Is Making AI-Generated Reports for Local Police
The CIA has been using artificial intelligence to write “open source intelligence” analyses, which are being distributed to local police via DHS “fusion centers,” according to documents obtained by Reason.
The CIA has been using artificial intelligence to write "open source intelligence" analyses, which are being distributed to local police via the Department of Homeland Security's "fusion centers," according to documents obtained by Reason.
CIA officials have been alluding to this new AI tool. Open Source Enterprise (OSE) Director Randy Nixon told Bloomberg in a 2023 interview that the agency was building a new chatbot for intelligence analysts to use, and CIA Deputy Director Michael Ellis said at an April 2026 conference that it had produced an entirely AI-generated report for the first time.
The use of AI has clearly grown quickly since then. OSE was creating daily reports on the security situation in Canada and Mexico during this year's soccer World Cup. Reason obtained the reports during an unrelated request under the Washington Public Records Act for World Cup–related police bulletins.
The reports don't contain anything secret or particularly shocking. They are an example of "open source intelligence," a longstanding practice of combing through public media for potential indicators. But they are a rare glimpse into how the CIA uses new technologies—and how U.S. intelligence agencies are integrated into local policing.
"OSE used a keyword search and a keyword relevancy algorithm to select these articles from major outlets in host countries. OSE used [generative AI] to refine the selection, translate the titles, summarize the selected articles. OSE officers reviewed and validated the GAI outputs at each stage to ensure they accurately reflected source material upon initial setup of this serial report," the methodology note on each of the reports states.
The note names GPT-4.1—which is also commercially available to the public—as the particular generative AI model used for these reports. In other words, CIA analysts are plugging international news stories into ChatGPT and passing on the results.
The reports were passed on to the Washington State Fusion Center, an office for sharing intelligence between federal, state, and local law enforcement. After the attacks of September 11, 2001, the new Department of Homeland Security created the national fusion center network, which has often been criticized for wasting resources and spreading downright incorrect information.
It's not clear what, if anything, police in Washington state did with the CIA reports. The fusion center redacted several of the documents it provided to Reason for security reasons.
The U.S. government has been increasingly integrating AI into the national security apparatus. Immigration and Customs Enforcement (ICE) has been using AI tools for facial recognition and real-time tracking of individuals, based on massive amounts of data vacuumed up from other government agencies and private industry. The FBI recently put out a request for contractors to build an AI software for maintaining watch lists, as Reason first reported.
Most consequentially, the military uses AI targeting systems, such as Project Maven, to mark enemy targets faster than a human being could. After the Minab school bombing, in which the U.S. military killed dozens of Iranian schoolchildren, there was early speculation that the school was targeted by a mistaken AI program. However, the attack turned out to be a likely case of human error instead.
As the fusion center documents show, artificial intelligence is making its way into national security work in much more mundane ways. And it's letting the government do more of what it already wanted to do in the first place—in this case, filling fusion centers with paperwork.
