Michael Dickens, writing on LessWrong, proposes that a frontier AI company should voluntarily shut down—not out of economic necessity, but as a deliberate signal to governments and competitors that artificial superintelligence poses genuine existential risk. Dickens argues this move would have outsized policy impact: a multi-billion-dollar company publicly concluding it can't make this technology safe would break the current coordination deadlock where each lab fears unilateral slowdown will let competitors race ahead. Shutting down sends a signal that coordination is possible. As an alternative, Dickens suggests reallocating an entire budget to safety research and global coordination efforts rather than frontier model development. He addresses practical concerns: that investors might sue and that shutting down would eliminate safety research capacity. But Dickens contends the current strategy—racing ahead while doing a dash of safety research on the side—is failing by its own logic. The essay frames shutdown as either a worthwhile political move or a thought experiment that exposes the tension between competitive pressures and actual safety.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton