The HUMIT & SIGINT Paradox

Jadon Teh

Prophet Enhanced Signals Processing Vehicles (ESP) Source: General Dynamics Mission Systems

"Intelligence is fundamentally a human endeavor," as CIA Director William Burns once affirmed. Human intelligence represents the essential elements that are not at risk in today’s intelligence community. Built over decades of Cold War experience, HUMINT relies on foreign assets, undercover officers, and building relationships that technology cannot replace. Agencies like the CIA and DIA use human resources to gather information that satellites and intercepts cannot obtain, such as the intentions of secretive regimes, the discussions of enemy leaders, and the personal weaknesses of foreign decision-makers. As Signals Intelligence (SIGINT) and commercial surveillance grow increasingly widespread, the central challenge may shift from technological capability to whether agencies retain the skills needed to manage human networks. As intelligence work moves from collecting information, to automated analysis, to machine-generated reports, each step makes the slow, expensive, and complex process of handling human resources seem less necessary. This can lead to dangerous gaps in understanding.

The key question is what happens when a human-source network breaks down, and adversaries take advantage of the gap. For instance, if the CIA’s Directorate of Operations cuts back its Non-Official Cover Program to focus more on cyber collection, and a secretive government makes a move that SIGNT cannot detect, it becomes unclear who is responsible for the intelligence failure. Is it the officer whose job was cut, the director who shifted the budget, or the policymaker who chose speed and savings? This is not just a theoretical problem. Before 9/11, intelligence failures were mainly attributed to problems with HUMINT, as opposed to weak SIGNT. The U.S. had little access to al-Qaeda leaders and could not answer the basic question: "What does the enemy plan to do?” The 9/11 Commission report found that a lack of human sources in terrorist networks was a major weakness that technology could not fix. Such a weakness could lead to further devastation if a nuclear-armed country came into the equation. When HUMINT falters, decision-makers lose the ability to understand the enemy's intent; not just enemy capability signals can reveal where a missile is, but a well-placed human source can reveal whether it will be fired. The costs of getting this wrong are not abstract; history has shown repeatedly that intelligence gaps at the human level produce failures no technology can compensate for. The threat environment today, however, has grown far more complex than any single attack. A similar gap now, particularly involving a nuclear-armed adversary or a state actor operating through proxies, could produce consequences that are not recoverable.

            HUMINT is not outdated, and this is not an argument to limit signal intelligence or AI analytics—the real issue is whether these tools are being used to support human collection or replace it. International law makes things even more complicated. HUMINT operations in foreign countries fall into a gray area: recruiting foreign nationals, sending officers under diplomatic cover, or gathering information within an ally’s institution falls between accepted espionage and actions that could be seen as a violation of sovereignty. The Vienna Convention on Diplomatic Relations and Executive Order 1233 set the legal rules at home, but neither was intended for a time when AI can quickly find human-source candidates, speeding up HUMINT in ways they cannot keep up with. Getting this wrong would be costly; instead of waiting for problems, legal rules for HUMINT need to be updated for AI-assisted targeting, and agencies should be required to keep a minimum level of HUMINT as a safeguard against the blind spots that come with relying too much on signals intelligence before an adversary who still values human collection takes advantage.