The company says the world must “get ready” for powerful AIs that can exploit vulnerabilities in websites within the next few months.
Claude Mythos, Anthropic’s new AI model, spooked staff when it found previously unknown security flaws across major web browsers and operating systems. Anthropic has only handed it to select organisations while it tries to plug whatever else the thing can prise open.
Speaking before the gathered throngs at the Semafor World Economy, Anthropic co-founder and head of public benefit, Jack Clark, said Mythos is “not a special model” and that rivals will soon ship similar capabilities.
“There’ll be open-source models from China that have these capabilities,” Clark warned.
According to the Financial Times, the partial release has already rattled financial authorities, with the US Treasury secretary summoning leaders of US banks to talk through the risk to financial security. British banking regulators have been doing the same, while industry bosses mutter about the banking sector’s cyber risk.
Anthropic is talking to the Trump administration about Mythos while it scrapes with the Pentagon over guardrails on the military’s use of its tools.
“We have a narrow contracting dispute, but I don’t want that to get in the way of the fact that we care deeply about national security,” Clark said.
“Our position is the government has to know about this stuff … So absolutely, we talk to them about Mythos, and we’ll talk to them about the next models as well,” he said.
