CIA’s Future Relies on Human-AI Collaboration, CAIO Says

IT officials must understand “how humans and AI are working together” to ensure responsible AI use in intelligence and security, CIA Chief Artificial Intelligence Officer Lakshmi Raman said on Tuesday during the AWS Summit in Washington, D.C.

“AI is able to speed up the processing to be able to do automation,” Raman added. “But, at the end of the day, who is taking on the risk or deciding the intent and making the decisions?”

Raman added that AI can significantly accelerate data processing and automate tasks, but ultimately, the critical decisions, risk assessments and determinations of intent remain firmly in human hands. While AI can enhance workforce efficiency and decision-making, having a trained and well-developed workforce is critical for the CIA’s mission.

“It’s about the human. Who is deciding the level of oversight, the risk to take and even whether to continue,” added Raman.

Leveraging AI to Streamline Accreditation Processes

Raman explained how AI and machine learning are being deployed to streamline the accreditation and authorization processes for CIA systems. These processes, she said, can be augmented and sped by automation and AI systems, especially in an era where technological advancements are moving at an unprecedented pace.

“We certainly don’t want to be slowed down by an accreditation process,” Raman noted. “We also still need to be safe. How do we do our best to keep the artificial intelligence and machine learning helping that along, while having the humans in the space.”

Enabling Security Teams to Work Faster

AI is instrumental in managing the “vast amounts of data” that flow into intelligence agencies at an exponential rate, Raman said. AI-powered tools assist in triaging this influx, rapidly sifting through information to surface insights and identify patterns that would be overwhelmingly time-consuming for human analysts alone. Raman cited the example of cybersecurity analysts, who traditionally manually review network data to block suspicious IP addresses.

“There’s an opportunity [for AI] to all of that and let the security analysists [make highl-level decisions],” she explained.

Keeping Humans in the Loop

The nascent state of AI systems and evolution of the technology means that answers may not always line up, Raman said. The inherent non-deterministic nature of AI models, where the same question posed multiple times can yield different answers, presents challenges to agencies like CIA. Raman said that this underscores the critical need for human judgment at the final stage of AI-driven processes.

“Having a human who can make a judgment at the end of what the AI comes up with is critically important,” Raman asserted, especially when AI is poised to take actions like blocking a connection or preventing an occurrence.

The Rise of Agentic AI

Agentic AI presents opportunities for CIA, Raman said, including automation. She said she could see AI agents transforming complex, multi-step workflows and facilitating “tool calling” across diverse databases and foundational tools. But the “black box” nature of AI means that leadership needs to focus on explainability and trustworthiness – paramount concepts for the CIA.

“Being able to trust AI is an incredibly important concept,” she said. “Having the ability to explain that to [CIA staff] is very important. Legal and data compliance is huge.”

By strategically integrating AI into the human workforce, agencies like CIA can enhance efficiency, accelerate processes and better manage overwhelming data, Raman said. If human oversight and decision-making remain at the core of their operations, Raman noted, AI will increase mission delivery, even navigating high-stakes environments like intelligence and national security.

“I think the productivity gains that AI can provide, when incorporated into enterprise workflows … are enormous,” said Raman. “We too have seen measurable increases in productivity by leveraging AI into workloads. It’s just such an exciting time to see how this can really transform some of the way that we do our trade craft.”