The CIA uses secret AI technology to make computer programs that talk like foreign leaders and help their analysts understand how those leaders might act. The New York Times says CIA analysts can use this new system to talk with computer-generated versions of world leaders, trying to guess how they might react in different situations.
For many years, CIA analysts have gathered information about world leaders by reading public records and analyzing secret reports. The CIA is using new AI technology called large language models to make their leader analysis work faster and better. The LLMs can answer questions by looking at huge amounts of information, making it easier for analysts to talk with digital copies of leaders instead of building profiles from static documents.
The CIA keeps secret how they created and used this chatbot, which matches their usual preference for keeping information private. As the CIA’s first Chief Technology Officer, Nand Mulchandani—who worked in Silicon Valley and led AI projects at the Pentagon—helps the agency find new ways to work. Since becoming Chief Technology Officer, he has worked to make it easier for the agency to develop new technologies compared to China’s faster pace.

Juliane Gallina, CIA’s digital innovation deputy director, says agencies must show what they’re working on to get private companies to want to help. The CIA wants to work better with tech companies by sharing some of its technology secrets. They believe this will make it easier for tech firms to get through the agency’s strict approval process.
The agency hasn’t revealed how much public money supports its AI projects, even after trying to be more open. The CIA’s use of AI is a big improvement, but it also makes people wonder how right or wrong it is to create digital versions of real people and whether the agency tells the public enough about how it works.