ControlAI Aims to Block Superintelligence in Business

Connor Leahy, executive director of the American NGO ControlAI, argues that current security methods are no longer sufficient for AI systems that surpass human capabilities. He advocates for preventing companies from aiming for superintelligence. He also claims that a "wave of legislation" is beginning to support this shift, an idea that seemed far-fetched to him just six months ago.

Leahy invokes a legislative dynamic and rejects alignment alone

Connor Leahy asserts that what seemed impractical six months ago is now receiving support from a wave of new legislation. In this context, ControlAI advocates for preventing companies from developing superintelligence. Leahy believes that the associated risks can no longer be managed solely through alignment. He also argues that containment, taken alone, is insufficient.

Incidents raise alarms about systems more capable than humans

A data breach involving OpenAI and Hugging Face has highlighted security incidents that illustrate the potential dangers of deploying AI systems that are more capable than humans. Meanwhile, AI companies present superintelligence as inevitable.

Who is behind this stance: the profile of the speaker and the NGO

Connor Leahy is both an artificial intelligence researcher and an entrepreneur. He leads ControlAI, a nonprofit organization based in the United States. He articulated these positions during an appearance alongside Rebecca Bellan on a podcast.