Should the United States act defensively in cyberspace? Dr. Michael Sulmeyer, Professor of the Practice at Georgetown University’s School of Foreign Service and the first Senate-confirmed Assistant Secretary of Defense for Cyber Policy, gives us a short answer: “We should be playing offense.”
“My goal, I would submit, is not deterrence,” he says in this episode of the Responsible AI podcast. He doubts that flaunting cyber capabilities talks anyone out of anything, because cyber operations pay off for the attackers. (North Korea funds its weapons program this way.) He says the U.S. should aim to disrupt and degrade how adversaries operate, making their work slower, harder, and costlier, enough “to throw sand in the gears.”
That strategy starts with vocabulary. Michael says the word “war” boxes in our thinking. Most cyber incidents land below the legal bar for armed conflict, so calling them war leaves policymakers with few good responses. He will take competition, conflict, or a “street fight,” as long as the words leave room to act. AI adds urgency. Tools now turn up software vulnerabilities by the boatload, and the tools for fixing them run behind. Michael sees promise in helping small businesses shore up the basics, starting with the eternal question of “is your password ‘PASSWORD’?”
The conversation then turns to China. In a Foreign Affairs article, Michael discusses what happens when China gets its own Mythos, a model with Anthropic’s Claude Mythos-level skill at finding flaws in code. He offers a blunt update: “It’s probably worth betting that they’ve got it at this point.” States keep these capabilities secret, so by the time the United States confirms it, China may have used the model for a while. The sharpest risk sits in critical infrastructure. Power and water systems run on equipment with decades-long replacement windows, and AI exacerbates the years’ worth of technical debt.
The final stretch covers humans and machines. Michael argues that instead of handing the building of AI over to AI itself, society should decide in advance where people stay in the loop. He closes with an invitation to you, the listener, especially mid-career professionals, to spend a couple of years in public service, where the need is for “matured judgment to make hard calls.”
Articles mentioned:
When China gets its own mythos: Preparing for an AI cyber crisis. Foreign Affairs. July 22, 2026
Artificial Intelligence and the Risk of Strategic Cyberattack. RAND. July 14, 2026
Topics
1:36 Are we in a cyber war?
3:43 Strategic vs. tactical cyber attacks
5:41 How AI is accelerating cyber conflict
8:53 Attribution: espionage or attack?
12:02 Government coordination and the Pentagon
14:30 "Layered effects" and the Venezuela operation
17:07 China, AI safety and crisis talks
19:49 "When China Gets Its Own Mythos"
22:06 Critical infrastructure and technical debt
23:57 Idaho National Lab's role in defense
26:05 Playing offense vs. deterrence
30:07 Who has the edge in the AI cyber race?
32:10 Ten years out: will AI make us more secure?
33:52 Humans in the loop and recursive self-improvement
35:49 AI in war games and nuclear modeling
39:53 A call to public service
About Professor Michael Sulmeyer: Dr. Michael Sulmeyer is Professor of the Practice at Georgetown University's School of Foreign Service. Previously, he was the first Senate-confirmed Assistant Secretary of Defense for Cyber Policy and the Principal Cyber Advisor to the Secretary of Defense. Prior to his appointment as Assistant Secretary, he served as the Principal Cyber Advisor to the Secretary of the Army. Before his time with the Army, he served in multiple positions in the Office of the Secretary of Defense.
Outside of government, Michael was the Director of the Cybersecurity Project at the Harvard Kennedy School’s Belfer Center for Science and International Affairs. He also taught at the University of Texas School of Law and was a Senior Fellow at Georgetown University’s Center for Security and Emerging Technology.
Listen: Apple Podcasts | Spotify | YouTube | RSS