AI Researcher Calls for Halt to Superintelligent AI Development
ControlAI's Connor Leahy argued on TechCrunch's Equity podcast that superintelligent AI development should be stopped entirely, citing safety incidents like OpenAI's Hugging Face breach as evidence such systems can become unmanageable. He noted this once-fringe view is now gaining traction through new legislation.
On a recent episode of TechCrunch's Equity podcast, AI researcher and entrepreneur Connor Leahy, now U.S. Executive Director of the nonprofit ControlAI, argued that companies should be stopped from developing superintelligent AI altogether, rather than attempting to make such systems safe through control measures.
Speaking with host Rebecca Bellan, Leahy referenced recent safety incidents, including OpenAI's data breach involving Hugging Face, as evidence of the dangers posed by deploying AI systems more capable than humans.
Control Is No Longer Enough
Leahy said the risks tied to systems humans can no longer reliably control have grown too large to be managed through alignment and containment techniques alone. In his view, the only responsible path forward is preventing companies from building such systems in the first place.
He noted that this position, which seemed far-fetched just six months ago, is now being reinforced by a wave of new legislation, with lawmakers increasingly moving to address the risks of advanced AI development.
ControlAI, the organization Leahy represents in the U.S., advocates for a far more radical approach to AI safety than most of the industry, pushing for an outright halt to superintelligent AI development rather than relying solely on technical safeguards.
