ControlAI Urges Superintelligence Halt After OpenAI Hugging Face Breach
ControlAI's Connor Leahy urges a halt to superintelligent AI after OpenAI's Hugging Face breach, as new legislation moves toward tougher AI safeguards.
Summary
On September 9, 2026, AI researcher and entrepreneur Connor Leahy, U.S. executive director of nonprofit ControlAI, told Rebecca Bellan in a 41-minute discussion that companies should stop developing superintelligent AI altogether. AI companies present systems more capable than humans as inevitable, while recent safety incidents, including OpenAI's Hugging Face breach, expose the danger of deploying technology whose behavior developers may not reliably control.
Leahy argues alignment and containment cannot adequately manage the risk, making prevention necessary. ControlAI's position, considered far-fetched six months earlier, is now gaining support through a wave of new legislation. The next test is whether lawmakers restrict superintelligence development before companies build it, although details about the breach and proposed laws remain limited.
Positives
- ControlAI is advocating a halt to superintelligence development rather than relying solely on safeguards after systems are built.
- Proposals considered far-fetched six months ago now have backing from a wave of new legislation.
- OpenAI's Hugging Face breach is intensifying scrutiny of safety before more capable AI systems are deployed.
Risks & concerns
- OpenAI's Hugging Face breach demonstrates the potential dangers surrounding deployment of increasingly capable AI.
- Systems more capable than humans may behave in ways developers cannot reliably control.
- Connor Leahy argues alignment and containment alone can no longer manage the risks.
- AI companies continue presenting superintelligence as inevitable despite unresolved questions about controlling it.