Bernie Sanders proposes global ban on superintelligent AI

Sanders is the first major US politician to say out loud what many alignment researchers have been writing in white papers: if you really believe superintelligent AI could be uncontrollable, the default policy should be to ban it, not manage it. His proposal to halt advanced AI development and prohibit superintelligence is unlikely to become law in its current form, but it moves the Overton window. For the big labs, this is a signal that the political system is starting to absorb arguments that once lived only in technical blogs and think-tank memos.

In the near term, the bill creates more scrutiny for companies launching models like Astra that flirt with AGI branding while admitting they are hard to monitor. It also gives activists pushing for compute caps and international treaties a high-profile ally. However, the US is in a strategic race with China and other powers, and many policymakers will view a unilateral ban as self-sabotage. That tension between existential risk and geopolitical rivalry is now out in the open. Even if the legislation stalls, expect future AI bills to borrow its language on “superintelligence” and to demand more transparency about rogue-agent incidents.