The CIA is preparing for a powerful artificial intelligence “game changer” to emerge and upend the spy business.
CIA Deputy Director Michael Ellis told The Washington Times it is “still unclear” when artificial general intelligence may arrive, but the agency is gearing up.
Top researchers and technologists around the world are striving to achieve AGI, meaning the development of AI systems that perform at least as well as humans across all cognitive domains.
Mr. Ellis said such new technology could particularly affect defenders and attackers in cyberspace and drones used in warfare.
“I think it does have a potential to be a true game changer, if it emerges, and one that will fundamentally reshape the intelligence business but national security more broadly,” Mr. Ellis said in an interview.
Key to the CIA’s preparations for advanced AI is preventing technological surprise from disrupting America’s national security.
China’s government and industry are racing toward AGI and pursuing superintelligence, which theoretically could outperform humans.
The communist country’s efforts for such technological advancement include traditional pathways and far more unorthodox strategies.
While China is working to improve the size and scale of its large language models akin to the AI efforts of American Big Tech companies, China is also investing in efforts to create a cognitive merger that physically combines man and machine.
Such pathways involve the non-therapeutic use of brain-computer interfaces, which link brains’ electrical signals to computing power, and embodiment, including through the deployment of powerful AI in robots.
China is testing its theories, including via a pilot program in Wuhan intended to spread throughout the country, according to a presentation from expert analysts at Georgetown’s Center for Security and Emerging Technology.
Many details on Chinese researchers’ efforts to pull ahead in the AI race are hidden from public view, however.
The arrival of a competitive and seemingly cost-effective model from China’s DeepSeek caught its American competitors at OpenAI by surprise earlier this year and U.S. companies failed to anticipate subsequent developments by Chinese researchers and technologists too.
The CIA, meanwhile, is leaving no stone unturned in its efforts to understand and harness AGI.
“I can’t get into the specifics of exactly what we’re looking into or working on in terms of those alternate approaches but I will say that we do take a very broad look at this topic,” Mr. Ellis said. “And our in-house [research and development] folks — we’ve got our [Directorate] of Science and Technology, we’ve got our CIA Labs, you look at In-Q-Tel making investments — we’re taking a broad look at all of these different approaches.”
The AI overhaul of the CIA is already underway as well. The agency is determined not to fall behind in its adoption of new technology.
“It’s transforming the way everyone at CIA works,” Mr. Ellis said. “And if it hasn’t transformed the way they work yet, it will be soon. And we’ll be pushing them, pushing all our workforce to integrate AI tools into how they do their daily jobs.”
Mr. Ellis said AI is rewriting the intelligence playbook, including by allowing its analysts to digest large volumes of information, speed translations and help operations officers plan their activities.
The agency also has its own AI solutions and is using them on the secret side of its work.
“We have in-house capabilities, we are bringing models onto the high side,” Mr. Ellis said. “If you aren’t using those [AI] models in your daily work, then you are not going to be successful at CIA.”
Mr. Ellis said the intelligence community has historically not monitored technological and economic targets effectively enough, but President Trump and CIA Director John Ratcliffe are looking to change that with a cultural shift.
Mr. Trump unveiled his administration’s new AI Action Plan last week, which is designed to ensure American dominance in the AI sector through an overhaul of federal policy.
Mr. Ellis said Mr. Ratcliffe is adamant that the agency get back to its core mission of aggressive operations and stealing secrets.
“This is going to be a race between us and the Chinese to see who can build, protect [and] govern AI faster,” Mr. Ellis said. “And this action plan is going to set us up, it’s going to set the government up, it’s going to set the American private sector up to prevail in that race.”
• Ryan Lovelace can be reached at rlovelace@washingtontimes.com.