Jensen Huang, CEO of the $5 trillion chipmaker Nvidia, told CBS News there is a "0% chance" AI destroys the world by 2030, calling former Anthropic researcher Jacob Coxon's extinction claims "irresponsible." The remarks land in the middle of an intensifying global debate over AI safety triggered by Coxon and two other Anthropic researchers who warned AI could become superhuman within the decade. Huang's most interesting move wasn't the dismissal of doomsday scenarios — it was his reframing of the regulatory question. He argued that AI companies calling for new safety regulation are actually seeking relief from existing laws covering cybersecurity, product liability, and contractual obligations. "Go and read between the lines," Huang said. "They're actually not asking for more laws. They're asking to be relieved of the laws we do have." This is a sharp structural observation wrapped in self-interest: Nvidia sells the shovels, not the gold, and existing product liability law hits the model deployers harder than the chip suppliers. The timing is crowded. Anthropic's own threat intelligence report details how criminals and state-sponsored actors have attempted to use Claude to design weapons, create pathogens, and surveil dissidents. Dario Amodei, Sam Altman, and Elon Musk have all called for slowing development. Donald Trump has dismissed AI anxiety as a "hoax" and framed slowdowns as ceding advantage to China. Treasury Secretary Scott Bessent announced a US-China AI dialogue mechanism to share safety alerts. The Tony Blair Institute added another layer, publishing a paper warning that countries failing to embrace AI face "profound and compounding disadvantage" — while simultaneously noting that state institutions are "simply not equipped" to handle the choices ahead. Blair himself wrote that AI will transform the world "within a decade," a timeline that sits uncomfortably close to the doomsday framing Huang dismisses. Huang's position contains a real insight and a real blind spot. The insight: bespoke AI regulation could indeed create a moat for incumbents while exempting them from liability regimes that already exist. The blind spot: existing law was written for products with predictable failure modes, not systems whose capabilities emerge unpredictably at scale. Cybersecurity statutes don't cover a model that autonomously discovers a novel pathogen synthesis pathway. The deeper extraction question is who benefits from each framing. Doomsday narratives that demand new regulatory bodies favor large incumbents who can afford compliance overhead. "Just apply existing law" favors the hardware layer — Nvidia — which faces less direct product liability than model deployers. Neither framing centers the public, which bears the cost of both under-regulation and regulatory capture. What's missing from every actor's position is a credible enforcement mechanism. The US-China dialogue is a start, but safety alerts between geopolitical rivals competing for AI supremacy is a thin reed. The TBI paper correctly identifies that institutional capacity is the binding constraint — not the absence of alarm, but the absence of competent response.
