IST's New Perry-Shultz Center Targets AI, Nuclear Risk and Strategic Stability

A wider definition of strategic stability

According to IST, the Center was launched in September, alongside a second new Center, and aims to shape technology's trajectory into a force for stability. It also seeks to build common cause among states and nonstate actors against the most consequential risks.

Steve Kelly said the term emerged from the Cold War nuclear standoff between the United States and the Soviet Union, when mutually assured destruction held the peace. He argues that this definition is no longer enough. Strategic stability still depends mainly on deterrence among states, but the conditions that make deterrence work can be shaped and eroded by technology-enabled dynamics at the societal level.

Kelly pointed to AI as the clearest example. He said it increasingly shows the ability to enable information warfare that erodes public trust at scale, and automated cyber attacks that could put critical infrastructure services at risk. Such effects become a strategic stability issue when they undermine a state's ability to mobilize and project power, which alters the deterrence calculus between states. Kelly also noted that Presidents Trump and Xi agreed in May to pursue a "constructive relationship of strategic stability."

Four projects and a lesson from 1979

The Center is named after former Defense Secretary William Perry and the late Secretary of State George Shultz. Sylvia Mishra recalled that in 1979 Perry, then Under Secretary of Defense, was warned in the middle of the night that Soviet missiles were headed for the United States. The warning was false and had been triggered by a training tape. Mishra said the lesson was that technical literacy, combined with human judgment and diplomacy, can prevent a technical failure from becoming a catastrophe.

She described four lines of work:

• AI-NC3: recommendations for policymakers on integrating AI into nuclear systems safely.
• CATALINK: a blueprint for a multilateral nuclear crisis communications channel.
• AI-Nuclear Policy Accelerator: AI literacy for a cadre of mid-career nuclear weapons and national security experts.
• U.S.-China work: managing AI-nuclear risks and the dynamics of the AI race, with the aim of creating conditions for strategic stability.

Epistemic capture and overreliance on data

Brandon Cortino, who works on the intersection of AI with nuclear command, control, and communications, said very few practitioners can discuss both the AI and the NC3 sides. He highlighted the concept of "epistemic capture," in which an AI system curates what data it shares in a way that changes how people understand the world. He compared it to a friend choosing one milk brand over the phone without mentioning the alternatives, then asked readers to imagine the same dynamic with a missile launch notification under extreme time pressure.

Cortino also said popular images such as Skynet miss the point. In his view, the biggest risk is not an evil robot uprising but human beings causing unintentional nuclear escalation by over-relying on poorly integrated data streams.

Space, 3D printing and undersea drones

The team also named technologies beyond AI that deserve attention. Catherine Murphy pointed to counterspace capabilities. Nuclear-armed states rely on satellites for command and control and crisis communications, and commercial and military space systems are increasingly entangled, so an attack on one satellite could hit another system. She noted that few international rules govern interference with space assets.

Larissa Alves Lozano highlighted additive manufacturing, or 3D printing, which she said has high dual-use capabilities and lacks regulation. She said it blurs the line between tangible and intangible export controlled goods and lowers the barrier to restricted equipment for non-state actors and rogue states under sanctions.

Mishra said she is optimistic about AI improving intelligence and situational awareness. Her concern is undersea warfare: increasingly networked underwater drones could detect, track and hunt nuclear-armed submarines, which could affect second-strike capabilities if their survivability can no longer be taken for granted.

What it means

The Q&A shows the Center treating AI as a factor in nuclear risk, information integrity and crisis communication, rather than a standalone technology issue. Its stated aim is to give policymakers practical recommendations and to build expertise among the people who work on nuclear and national security questions. Even the Center's purple branding reflects the approach, according to Kelly: a blend of the red of nuclear danger and the blue of diplomacy, reflecting the deterrence and trust-building traditions of its two namesakes.