The debate over Jensen Huang’s “don’t ship” approach to AI safety

Semafor reports that Huang dismissed fears of AI’s existential threats as a “distraction,” while arguing that labs unable to contain their models should close. ZeroHedge reports that he endorsed third-party safety auditors and more computing power for evaluation and alignment, but rejected the idea that competitive pressure forces labs to ship unsafe products. A critical essay argues that this misses the problem: companies have incentives to race ahead, while the costs of harm can be delayed, spread across society and poorly addressed by ordinary litigation.