The CIA is planning for a future in which artificial intelligence agents augment the spy business, but a human stays involved in the process.
Lakshmi Raman, the CIA’s AI chief, said Friday that she anticipates AI agents will be heavily involved in the espionage business.
“I can imagine a future where AI is in the hands of an intelligence, of all intelligence officers, helping them do their work faster and have lots of productivity gains,” Ms. Raman said at a Vanderbilt University summit. “I can imagine having AI assistants in place to help our analysts go through massive amounts of reporting from thousands of sources with all modalities.”
The CIA has already deployed generative AI tools on its top secret systems and made various services available across the entire intelligence community, Ms. Raman told attendees of Vanderbilt’s Summit on Modern Conflict and Emerging Technology.
Ms. Raman said the CIA is spending a lot of time exploring, developing and deploying AI tools amid the generative AI boom in recent years, amid the arrival of commercial tools such as OpenAI’s ChatGPT.
“We have seen a lot of work inside on our top secret fabric, around areas like leveraging generative AI for summarization, for search and discovery aids, for ideation aids, for our analysts to help them think about generating counter-assumptions to their arguments,” she said.
CIA’s Open Source Enterprise has also deployed a tool available to the entire intelligence community that helps analysts and officers benefit from commercially available information much faster than they could manually, according to Ms. Raman.
Such tools are meant to augment human intelligence officers and analysts, and Ms. Raman said the spy agency is carefully adopting and deploying the cutting-edge technology.
“We understand when it’s important to have a human in the loop, to apply human-machine teaming, to ensure that our AI aligns with our mission, our laws, and our intelligence community standards,” she said. “The fact is that I see AI not just as an emerging technology but as a strategic necessity.”
Ms. Raman said the CIA is also watching for the arrival of artificial general intelligence, anticipated AI systems that surpass human abilities, but she said it is “still unclear when it’s going to hit” and what the consequences are for national security.
Vanderbilt University Vice Provost Padma Raghavan pressed Ms. Raman for her view on potential threats from such advanced AI.
Ms. Raghavan asked about the potential for AI to function as “sleeper agents” akin to the human spies dramatized in “The Americans” television show that depicted Russian spies hiding in suburban Washington, D.C.
“We don’t know,” Ms. Raman said. “That’s why it’s a known unknown.”
