The US vice-president JD Vance has rejected calls from AI companies for global regulation of the technology's safety risks, telling firms building advanced models: "If you're building Frankenstein, stop." Speaking at an AI summit in Los Angeles, his remarks were widely seen as directed at Anthropic co-founder Dario Amodei, who has urged Washington to coordinate international oversight of AI, including with China. The intervention comes amid growing global unease over the potential dangers of artificial superintelligence, and highlights a split within the US administration, with President Trump dismissing fears of AI destroying humanity as a "HOAX".

The row follows warnings from Anthropic figures, including Amodei's claim that AI agents could be capable of "taking over the entire internet" within six to 12 months, and alignment lead Evan Hubinger's suggestion of a greater-than-10% chance AI could kill all humans within a decade. Vance, who suspects such warnings are a "Trojan horse" for regulatory capture, said companies should look inward rather than seek government rules. Meanwhile OpenAI, which lost control of AI agents that attempted to hack Hugging Face this summer, said it backs bipartisan US efforts on catastrophic AI risks, while Meta's Mark Zuckerberg said his firm had unilaterally delayed its Muse model for safety reasons without waiting for rivals to act.