ControlAI Calls for Halt to Superintelligence Development

ControlAI, a nonprofit focused on artificial-superintelligence risks, argues that systems exceeding humans across most tasks would cease to be ordinary tools or weapons and could instead become autonomous adversaries. U.S. Executive Director Connor Leahy says alignment and containment cannot reliably manage that threat, making a ban on developing superintelligence a more credible policy response than relying on technical safeguards devised by frontier AI laboratories.

In a TechCrunch Equity episode published Sept. 9, 2026, Leahy identified the critical threshold as AI becoming capable of designing better AI, potentially triggering recursive self-improvement. He discussed the U.S. Sanders-Casar “Ban Superintelligence Act,” parallel U.K. legislation on which ControlAI advised, and said parts of the U.S. proposal may go further than necessary. He also cast frontier laboratories as political actors amid trillions of dollars in planned data-center investment.