AI researcher calls for halting development of superintelligence
On TechCrunch's Equity podcast, researcher Connor Leahy argues that building superintelligent AI has become too risky and should be stopped altogether. His nonprofit ControlAI is now pushing this stance through new legislation.
AI companies have long framed superintelligent AI as an inevitable next step. But recent safety incidents, including OpenAI's data breach on Hugging Face, are highlighting the dangers of deploying systems whose capabilities exceed human ones.
On TechCrunch's Equity podcast, host Rebecca Bellan spoke with Connor Leahy, an AI researcher, entrepreneur, and recently appointed U.S. Executive Director of the nonprofit ControlAI. The organization advocates for a more radical approach to AI safety: preventing companies from developing superintelligence in the first place.
Why containment may no longer be enough
Leahy explains why he believes the risks posed by increasingly capable AI systems have grown too large to manage through alignment and containment methods alone. According to him, an idea that seemed far-fetched just six months ago — an outright halt on superintelligence development — is now gaining traction through a wave of new proposed legislation.
The conversation centers on what happens when humans can no longer reliably control AI systems once those systems surpass human capabilities, a concern that has gained urgency following recent industry safety incidents.
The episode is part of Equity, TechCrunch's flagship podcast, available on YouTube, Apple Podcasts, Overcast, Spotify, and other platforms.
