The Ban That Treats Superintelligence Like a Warhead

Penalties borrowed from the bomb

Sanders’s office compares the criminal penalties to unlawful nuclear-weapons development: up to twenty years for people, and a corporate death penalty for entities that try to route around the pause. That is the bill’s philosophical tell. It treats the labs’ own extinction rhetoric as a category of state security, not a branding problem. Casar argued that pausing dangerous American development also slows China, because Chinese labs train from American models. Whether that chain holds is an empirical fight. The statute writes it as policy anyway.

The same afternoon, Altman told the Security Council that labs alone should not write AI’s rules, while warning against doomerism. Sanders is doing the opposite of soft language. ‘When you are racing towards a cliff, you don’t just ease up on the gas pedal. You hit the brakes,’ he said in the introduction statement. He wants Congress to treat AI as an existential threat before the next training cluster finishes.

Honest severity, thin coalition

Progressives do not own a unified AI line. Sen. Elizabeth Warren has backed a pause without endorsing a permanent superintelligence ban. Other Democrats want bilateral pacing deals with Beijing. Sanders and Casar chose the hard edge: ban the capability class, stand up a department, export the prohibition through alliances and controls. Multiple employees at frontier labs are reportedly endorsing the bill. That endorsement is easier to give as a staffer than to obey as a founder racing toward a trillion-dollar IPO.

The Ban Artificial Superintelligence Act will not pass tomorrow. It still does the useful work of taking the industry’s risk talk at face value. If Amodei can tell the Security Council that poorly managed AI endangers humanity, Congress can write a statute that sounds like arms control. The remaining question is whether anyone with a gavel will treat that sentence as law rather than as a TED Talk.