Vance Rejects Global AI Safety Rules in Clash With Anthropic Chief

US Vice President JD Vance dismissed calls for global regulation of AI safety risks at an AI summit in Los Angeles. The remarks, reported Sept. 16, were aimed at firms building the most advanced AI models and executives urging Washington to constrain them. The Guardian

Vance directed his comments to Dario Amodei, co-founder and chief executive of Anthropic. "If you're building Frankenstein, stop," Vance said. He added: "If you're gonna create Frankenstein, don't come to the government and say we need regulation." His argument was that developers should stop systems they see as uncontrollable rather than build them and seek state protection.

Amodei had urged Washington, D.C. to coordinate control of AI systems, including with China. He warned that a swarm of AI agents — software that can act online with little human direction — could be capable of taking over the entire internet in six to 12 months.

Amodei set out his case in a 3,800-word essay urging companies to slow improvement of their most advanced models. The New York Times He called for a global slowdown of artificial intelligence development. He said any slowdown by democratically governed countries should be limited to preserving the lead U.S. firms hold over China. Reuters

Vance rejected that sequence. He said if the cat is out of the bag, AI firms should build the defensive mechanism. He criticized AI leaders for seeking global regulation while withholding defensive tools. ANI He spoke on the All-In Podcast. On regulation, Vance said: "I feel a little weird." In February 2025, Vance told Europeans that massive regulations on artificial intelligence could strangle the technology. Reuters

The broader context here is a clash of two control theories. One says frontier capability crosses borders and needs enforceable coordination, even between rivals. The other says the American state is the wrong tool and responsibility sits with the labs that build the capability. It moves his February warning about broad restraint to a direct reply to lab-led calls for coordinated restraint.

Looking at what this means for diplomacy, the Washington-Beijing link is hardest to resolve. Amodei tries to pair restraint with competition by limiting it to what preserves the U.S. lead over China. Vance sets aside verification and says builders must supply defenses or not build. That leaves open who defines an unsafe system, who audits defense claims, and what leverage allies or multilateral bodies keep when the leading power declines a regulatory role.