ControlAI calls for global ban on superintelligent AI to avert extinction

Connor Leahy, the U.S. Executive Director of the nonprofit ControlAI, is calling for a total international prohibition on the development of artificial superintelligence (ASI). Leahy warns that such systems pose an existential threat to humanity and must be stopped before they are realized.

Leahy argues that ASI should not be viewed as a tool or a weapon, but rather as an "adversary." According to Leahy, it is not in the rational interest of any nation—specifically citing the U.S. and China—to pursue ASI because the resulting system would eventually control its creators. "When you build super intelligence, you don't have super intelligence," Leahy stated. "Super intelligence has you."

A Shift Beyond Alignment

This push for prohibition marks a significant departure from traditional AI safety discourse. While many researchers focus on "alignment"—the effort to ensure AI goals match human values—or "containment" strategies, ControlAI asserts that these methods are insufficient for a superintelligent entity. Leahy, who previously founded the AI safety startup Conjecture and the open-source LLM collective EleutherAI, now leads ControlAI in briefing lawmakers to move the policy framework from safety guidelines to absolute prohibition.

The Path to Deterrence

To enforce such a ban, ControlAI proposes deterrence measures modeled after international arms-control treaties. These measures would involve the strict tracking of the physical infrastructure required to train massive models, including the monitoring of GPUs, specialized chips, energy consumption, and data center infrastructure. By controlling the hardware, proponents argue that the global community can prevent the clandestine development of ASI.

Legislative Momentum

These views are already influencing policy discussions in major powers. Leahy has provided direct input on the "Ban Superintelligence Act" introduced by Sanders and Casar in the United States, as well as parallel legislative efforts currently being explored in the United Kingdom.

Why It Matters

If adopted, this approach would fundamentally transform the AI industry. It would shift the global regulatory landscape from a permissive environment of "responsible innovation" to a restrictive regime of hardware surveillance and international treaties. The movement reflects a growing belief among a subset of experts that superintelligence is inherently uncontrollable, making its existence a binary risk to human survival.

What's Next

Observers are now watching whether the "Ban Superintelligence Act" gains traction in the U.S. Congress and if the U.K. government will formalize similar prohibitions. The primary challenge remains international cooperation; for a ban to be effective, it would require a level of transparency and trust between geopolitical rivals that has rarely been achieved in the history of weapons proliferation.