Former U.S. Deputy Director of National Intelligence Karen Gibson
Dual-Use Dilemma of AI and Cyber Technologies
Building National Security Through Public-Private Partnership

“The private sector does take on an increasingly important role in national security. But it also would be a mistake to imagine that the government can regulate technology that it doesn't deeply understand without some kind of sustained partnership with the people who are developing it.”


Karen Gibson, former Deputy Director of the U.S. National Intelligence Service, is being interviewed by The Asia Business Daily at the Grand Hyatt Hotel in Seoul. Photo by Jo Yongjun

Karen Gibson, former Deputy Director of the U.S. National Intelligence Service, is being interviewed by The Asia Business Daily at the Grand Hyatt Hotel in Seoul. Photo by Jo Yongjun

View original image

Karen Gibson, a former U.S. deputy director of national intelligence, emphasized the dual-use nature of artificial intelligence and cyber technologies in an interview with The Asia Business Daily during her visit to Seoul for the ‘World Emerging Security Forum (WESF)’ on September 9, 2026.


A retired U.S. Army lieutenant general, Gibson served in the military for 33 years and held senior positions in intelligence and cyberspace. She served as a deputy director of national intelligence at the Office of the Director of National Intelligence (ODNI), director of intelligence at the U.S. Central Command, and deputy commanding general of the U.S. Army Cyber Command. After retiring from the military, she served as Sergeant at Arms of the U.S. Senate from 2021 to 2025. She is currently a Senior Associate (Non-Resident) at the Center for Strategic and International Studies (CSIS).


Gibson said AI could assist human decision-making, but that decisions in high-risk areas such as the employment of nuclear weapons must remain under human control. She also said traditional concepts of deterrence do not readily translate to cyberspace, where attribution is difficult and malicious actors include not only states but also criminal groups. Rather than trying to prevent all cyberattacks, she suggested South Korea should build resilient systems that can continue to accomplish their missions even when disrupted.


Karen Gibson, former Deputy Director of the U.S. National Intelligence Agency, is being interviewed by The Asia Business Daily at the Grand Hyatt Hotel in Seoul. Photo by Jo Yongjun

Karen Gibson, former Deputy Director of the U.S. National Intelligence Agency, is being interviewed by The Asia Business Daily at the Grand Hyatt Hotel in Seoul. Photo by Jo Yongjun

View original image

Q. Do you see private companies becoming national security actors?


A. Artificial intelligence and cyber have a dual-use nature. The government does not have a monopoly on developing these technologies.

Most of these capabilities are going to come out of the private sector because the government has some expertise and access problems. The private sector does take on an increasingly important role in national security. But it is unrealistic to expect private companies to make policy on how to protect these technologies. At the same time, it also would be a mistake to imagine that the government can regulate technology that it doesn't deeply understand without some kind of sustained partnership with the people who are developing it.

The responsibilities need to be divided. Industry should be responsible for knowing its technology, testing dangerous capabilities, securing its models, reporting incidents, and exposing them to greater evaluation. The government needs to understand the strategic systems that those technologies enter and establish boundaries. It must preserve human decision-making authority and decide what risks are acceptable as a society. It needs to be a partnership between the two.


Q. To what extent should AI be allowed to replace human judgment in military decision-making?


A. There are places where it is safe for a machine to make a decision. There are places where it is absolutely not safe for a machine to make the decision. There are three categories. In low-risk, low-consequence situations, machines can make decisions. Other situations require human supervision. A third category includes decisions such as the employment of nuclear weapons, which must remain under human control. Instead of making a machine the decision-maker, we should create human-machine teams in which the machine assists human decision-making.

In areas where the consequences are unacceptable, we need to build in safeguards that would prevent the machine or the artificial intelligence from acting independently.


Q. Is the use of nuclear weapons a clear red line?


A. That's a very clear red line that should absolutely be a human decision. If it's a dense urban area with many civilians and innocent non-combatants, that is not an area where I would allow the machine to make a decision. If it were a battlefield involving only combatants, perhaps that could be an area where a machine decision is acceptable. The important thing is identifying the areas where we will not permit the machine to make the decision.


Q. Should AI be allowed to make judgments involving the distinction between civilians and combatants?


A. There are many places where we should not allow AI to replace human judgments. There are times and places where it's okay for a machine to make a decision. If the consequences are minimal, there are situations where we can trust a machine. But there are other areas where I will not trust a machine. Those are areas where we must insist that human judgment remains. It starts with assessing the risks and consequences of various machine-based decisions.


Q. What concerns you most when AI, cyber capabilities and nuclear weapons converge?


A. If we're talking about the convergence of these three technologies, I would begin to have concerns if there were a scenario in which the data could become corrupted. We need to maintain a “healthy skepticism” regarding the answers that we may receive from artificial intelligence. There are two extremes to avoid. One is automation bias, believing what AI provides 100 percent and never questioning it. The other is automation rejection, not believing anything AI provides. We cannot abandon our role in making judgments entirely to artificial intelligence. We need to continue to exercise human judgment.


Q. In a conflict in the Taiwan Strait, do you think China would seek to disrupt commercial low-Earth-orbit satellite networks?


A. I think they would probably seek to. I don't know if it's the biggest concern, but it is something the Taiwanese should be aware of and work to address by building redundant, resilient capabilities.

Gibson said one advantage of widely available, low-cost systems is that they make it harder to completely deny communications capabilities.


Q. South Korea has recently seen cyberattacks targeting government networks, including systems linked to national security. The U.S. government has also disclosed attacks by hackers linked to China targeting government agencies and critical infrastructure. If advanced AI agents become widely used in state-sponsored cyber operations, will existing cyber deterrence strategies still be effective?


A. Cyber deterrence is something that many governments have struggled to develop. It's an imperfect analogy to think that the deterrent strategies we use for nuclear weapons apply in cyber.

Not everyone who conducts cyberattacks is a state actor. There are criminal elements and ransomware groups. There are also challenges of attribution in cyberspace— you can't always tell who conducted an attack, and that complicates deterrence. Malicious cyber activity occurs all the time. Nuclear exchanges, by contrast, are very catastrophic and very binary. We're subject to malicious cyber activities all the time. Concepts of deterrence in other realms don't translate readily to cyber.


Q. Does the difficulty of identifying an attacker make cyber deterrence less effective?


A. It certainly complicates things. It makes deterrence more challenging. If an attacker knows they will not be identified, it's not much of a deterrent.


Q. What cyber and AI security capabilities should South Korea strengthen most urgently?



A. It is not realistic to think that you can entirely prevent cyberattacks. That's an impossibly high bar. Rather than trying to prevent disruption entirely, South Korea should work to ensure that it has resilient systems where disruption does not entirely prevent them from accomplishing their mission. That requires a focus on redundancy, dispersion, alternate communication methods, proliferated space architecture, hardened facilities, distributed logistics, and rapid reconstitution. And let's not forget an ability to continue to operate in a degraded environment.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing