AI Won't End Nuclear Deterrence (Probably)

Deterrence limits technological domination

Nuclear deterrence prevents even economically and militarily superior states from imposing their will on rivals, creating a 'balance of nerves' rather than a pure balance of power.

Undermining deterrence raises coercion risks

If AI enabled a 'splendid first strike' or neutralized retaliation capabilities, technologically dominant states could threaten and coerce nuclear rivals with unprecedented impunity.

Survivability ensures mutual destruction

Secure second-strike capability—the ability to retaliate after absorbing a nuclear attack—depends on hiding forces across land, sea, and air to ensure no adversary could destroy everything at once.

Submarines are the most survivable leg

Nuclear-powered ballistic missile submarines represent the ultimate deterrent because they operate in tens of millions of square miles of ocean where electromagnetic radiation cannot penetrate and acoustic detection faces massive noise and distortion.

States use different survivability strategies

The US maintains a triad of land-based missiles, submarines, and bombers; the UK relies solely on submarines; while Russia and China use road-mobile launchers driving on highways, a tactic deemed politically unviable in the US.

Three pathways to erode deterrence

AI could theoretically enable a disarming first strike by locating all enemy weapons, disable nuclear command-and-control networks, or strengthen missile defenses enough to neutralize retaliation.

Sensor fusion faces physics barriers

While AI could fuse data from sonar, magnetic anomaly detectors, and satellite radar to track submarines, the ocean's vast volume, complex terrain, and increasing ambient noise create fundamental physical limits to transparency.

Move-countermove dynamics constrain AI

States will adapt to AI surveillance by engineering quieter submarines and deploying countermeasures, while deploying millions of sensors in contested waters risks sabotage and interference.