AI Safety Expert Calls for Global Ban on Superintelligence to Prevent Extinction

AI safety researcher Connor Leahy is calling for a total international prohibition on the development of artificial superintelligence (ASI), warning that such systems must be viewed as adversaries rather than tools. Speaking on Democracy Now!, the executive director of ControlAI argued that the pursuit of ASI poses an existential threat to humanity, necessitating a global ban to prevent human extinction.

Leahy first gained prominence in 2019 after reverse-engineering OpenAI's GPT-2 model. He now contends that the risks associated with superintelligent AI are too great to be managed by current safety protocols. His organization, ControlAI, is actively advocating for a comprehensive international treaty to halt the development of these systems. This push for a ban coincides with intensifying grassroots opposition; for example, retired teacher Wynd Kaufmyn was recently jailed following a protest and sit-in at OpenAI headquarters.

The Shift Toward Existential Risk

The discourse surrounding AI safety has evolved rapidly, moving beyond immediate concerns over algorithmic bias and workforce displacement toward the concept of "existential risk." This shift is supported by some of the field's most influential figures, including Nobel laureate Geoffrey Hinton. Hinton has warned that as AI systems evolve, they may eventually outthink their creators, rendering human oversight obsolete.

Regarding the difficulty of maintaining authority over advanced systems, Hinton stated, "I don’t believe we’re going to be able to keep control of them in the simple way of just outthinking them so they can’t escape."

Implications for Global Governance

The argument for a total ban suggests that existing regulatory efforts, such as moratoriums on data center expansion, are insufficient to address the scale of the threat. Critics of unrestricted AI development argue that if a superintelligent system is created without foolproof control mechanisms, it could act in direct opposition to human interests. In such a scenario, the resulting misalignment between AI goals and human survival could lead to global catastrophe.

The Path Forward

As advocacy groups like ControlAI increase pressure on governments, the central question remains whether the international community can coordinate a binding prohibition before a superintelligent system is realized. While some labs continue to push the boundaries of large-scale models, the growing consensus among safety skeptics is that the window for establishing meaningful control is closing. Observers are now watching to see if governments will move toward a formal treaty or continue to rely on voluntary safety commitments from the industry.