Vance Criticizes Anthropic: Don't Create a 'Frankenstein' and Then Seek Government Regulation

Beating AI News Flash: U.S. Vice President JD Vance sharply criticized the recent AI safety approach at the All-In Summit. He said that some frontier AI companies claim to have created "Frankenstein" while pointing to global AI governance as the solution. Vance argued that if a dangerous model has truly been created, the first step should be to stop; if the capabilities have already been released, then defensive tools should be built first.

He then turned his fire on Anthropic. Vance said Anthropic's latest model already possesses strong cyberattack capabilities, yet some enterprises urgently needing tools of the same level for defense cannot gain access. He did not specify which companies were denied, but Anthropic's Mythos 5.1 is indeed only open to a small number of vetted institutions, which the company says is because the model's capabilities in cybersecurity and biological research are too strong and pose a risk of misuse.

Vance also stressed that he does not believe Dario Amodei is deliberately manufacturing panic for regulatory capture. On the contrary, he said the assessments he has heard are that Dario is very serious and genuinely believes in these risks. What he really wants to criticize is Anthropic's approach: if it believes the model is dangerous, it should stop first, or hand defensive tools to those who need them, "don't build Frankenstein and then come to the government saying we need regulation."