CIA Deputy Director Michael Ellis: AI Is Changing the Speed of Intelligence
At the Billington CyberSecurity Summit, Ellis described a CIA reorganizing around cyber operations, emerging technology and a world in which intelligence advantage may increasingly depend on speed.
WASHINGTON, D.C. — The intelligence contest of the next decade may not be decided simply by which country possesses the most information. It may be decided by which country can understand and act on that information fastest.
That was one of the clearest messages from CIA Deputy Director Michael Ellis during the opening day of the 17th Annual Billington CyberSecurity Summit in Washington, D.C., where artificial intelligence, cyber operations and competition with China dominated his discussion of the changing intelligence landscape.
Ellis, who was sworn in as deputy director of the Central Intelligence Agency in February 2025, has spent much of his career at the intersection of intelligence, national security law and technology. His previous positions include general counsel of the National Security Agency, senior director for intelligence programs at the National Security Council and general counsel of the House Permanent Select Committee on Intelligence. He also served as an intelligence officer in the U.S. Navy Reserve and, outside government, as general counsel of the technology company Rumble. A graduate of Dartmouth College and Yale Law School, Ellis is also, somewhat unexpectedly, a former Jeopardy! champion.
Speaking at Billington, Ellis described an agency attempting to adapt to a technological environment moving much faster than traditional government institutions were designed to accommodate.
The problem is not merely that artificial intelligence gives adversaries new capabilities. AI changes the speed and scale at which intelligence itself can be collected, processed and exploited.
That distinction has enormous implications.
Intelligence at Machine Speed
For decades, intelligence agencies have faced the fundamental problem of finding meaningful signals within enormous amounts of information. AI radically expands their ability to do that.
Tasks that once required analysts hundreds of hours can increasingly be accomplished in hours. In cyber operations in particular, Ellis said AI provides a level of speed and scale that would otherwise be impossible.
But the same tools are available to America's adversaries.
China, Russia, Iran and North Korea remain major intelligence and cyber concerns, and AI allows operations to be conducted at a scale that would previously have required far greater human resources.
The result is something resembling a technological cat-and-mouse game.
Attackers need to succeed once. Defenders must succeed repeatedly.
And increasingly, the advantage belongs to whoever possesses the superior technology — and can deploy it first.
For the CIA, that means simply possessing cutting-edge technology is not enough. The Agency must also be capable of getting that technology into the hands of its officers quickly.
Ellis pointed to government acquisition as one of the obstacles.
Historically, he said, bringing new technology into the CIA could take an average of two to three years. In the contemporary technology environment, two years can represent several generations of capability.
The CIA has therefore established a target of approximately six months for acquiring and deploying new technologies. Ellis said the Agency has already completed more than 400 acquisitions within that timeframe.
That seemingly bureaucratic change may be one of the more consequential elements of the CIA's technological transformation.
A government that develops sophisticated AI strategies but cannot purchase or deploy technology quickly enough risks entering every technological contest several steps behind.
Reorganizing the CIA Around Cyber
Ellis also described organizational changes intended to move cyber capabilities closer to the center of intelligence operations.
Rather than treating cyber as a separate technical specialty supporting traditional intelligence work, the Agency is increasingly integrating cyber capabilities directly into intelligence collection and operations.
The CIA has created organizational structures designed to concentrate cyber intelligence and accelerate the delivery of technology to officers.
The underlying idea is straightforward: cyber is no longer merely something that supports intelligence.
Cyber is part of intelligence.
That shift reflects a much larger transformation underway across national security. Artificial intelligence, semiconductors, biotechnology and cyber capabilities increasingly overlap with traditional military, political and economic power.
The boundaries between those categories are disappearing.
The China Problem Is Different
China illustrates the difficulty particularly well.
Unlike the Soviet Union during the Cold War, China is simultaneously a strategic competitor and deeply intertwined with the American economy.
That creates an intelligence problem very different from the one Washington confronted during the twentieth century.
Military capabilities matter. Political intentions matter. But so do semiconductor supply chains, artificial intelligence research, biotechnology, private companies, investment flows and emerging technologies.
The targets themselves are changing.
Understanding a competitor like China therefore requires intelligence agencies to look beyond traditional government and military structures toward technological and economic ecosystems.
It also makes the private sector increasingly important.
American companies often possess information, technical expertise and visibility into emerging technologies that government agencies do not.
Ellis emphasized the importance of better leveraging that knowledge and treating private industry as a partner rather than simply a government contractor.
That idea appeared repeatedly throughout the Billington Summit: America's technological strength resides largely outside government.
The challenge is figuring out how to connect it to national security without destroying the very private-sector dynamism that created it.
The Risks Inside the AI
AI does not simply solve intelligence problems. It creates new ones.
Among them is data poisoning — the deliberate manipulation of information used by AI systems.
An intelligence service increasingly dependent upon AI-assisted analysis must be able to trust the information entering those systems. If an adversary can corrupt the underlying data, it may be possible to manipulate the conclusions produced from it.
That turns provenance, authentication and data integrity into national-security questions.
There is another danger as well.
The volume of modern intelligence may eventually become impossible for humans to analyze without AI assistance.
At that point, intelligence agencies face an uncomfortable dependency: they need AI to understand the information environment, while simultaneously needing to protect themselves against manipulation of the AI and the information feeding it.
The problem becomes recursive.
We need machines to help determine what is true, while adversaries increasingly use machines to manufacture convincing falsehoods.
Who Works at the CIA Is Changing Too
Technology also changes what an intelligence officer looks like.
Ellis discussed the need for more technical expertise within the Agency, including people with strong STEM backgrounds who are also motivated by the CIA's national-security mission.
The traditional image of intelligence work — spies, analysts, diplomats and linguists — is expanding.
Computer scientists, engineers, data specialists and AI researchers increasingly belong inside that picture.
Ellis has described the goal as maintaining a merit-based, mission-focused workforce. When he joined CIA leadership in 2025, he told employees that their work should remain free from politics and bias and emphasized building what he called a strong and capable intelligence service.
The workforce question may prove as important as the technology itself.
AI can identify patterns.
It can process enormous quantities of information.
It can accelerate analysis.
But intelligence ultimately exists to inform decisions made by human beings.
The Emerging Intelligence Equation
The Billington CyberSecurity Summit's 2026 theme is “Reducing Risk in An Age of AI-Enabled Threats,” bringing together more than 3,000 participants and hundreds of government and industry leaders in Washington from September 8–10.
Ellis's remarks captured why that theme extends far beyond cybersecurity.
Artificial intelligence is beginning to alter the entire intelligence cycle.
Collection becomes faster.
Analysis becomes larger.
Cyber operations become more integrated.
Technology acquisition must accelerate.
Private industry becomes more important.
And the distance between discovering something and acting upon it becomes shorter.
For most of the intelligence age, information itself was scarce and therefore powerful.
We may now be entering the opposite world.
Information is abundant.
Attention, verification and decision speed are scarce.
The intelligence advantage may therefore belong not to the country that collects the most data, but to the country that can move most effectively through a new chain:
DATA → TRUST → DECISION → ACTION
AI dramatically accelerates the distance between those points.
The challenge for the CIA — and ultimately for the United States — is ensuring that human judgment can keep up.
AI Transparency Note: This article was developed from the author's firsthand notes taken while attending the 2026 Billington CyberSecurity Summit in Washington, D.C. AI tools were used to assist with organization, background research, fact-checking and editing. The reporting observations and editorial direction are the author's.