ControlAI’s Connor Leahy on Why Superintelligence Is ‘Not a Weapon, It’s an Adversary’

In a recent discussion on AI risk, ControlAI U.S. Executive Director Connor Leahy argues that superintelligent AI should be conceptualized as an autonomous adversary rather than a controllable weapon or tool. Pointing to mounting safety breaches and coordination friction around frontier deployments, Leahy contends that human oversight cannot be taken for granted once capability surpasses biological limits. The dialogue emphasizes that managing highly capable models is increasingly a confrontation with unpredictable agency rather than standard software engineering.