The boss of artificial intelligence firm Anthropic, Dario Amodei, is expected in Washington DC on Monday for an emergency meeting with US Commerce Secretary Howard Lutnick — summoned days after the company was forced to block its just-released AI tool from public access because of a suspected national security vulnerability.
Anthropic blocked all public access to Fable 5 on Friday, a version of its Claude Mythos AI model that the company itself had said was "too powerful" for general release. The US government had prohibited Anthropic from allowing any foreign national access to the technology, after becoming aware of a potential "jailbreak" — an opening that could allow someone to make the tool do something it was not designed to do. Anthropic said it had received only "verbal evidence" of the purported vulnerability.
“Anthropic CEO Dario Amodei will meet US Commerce Secretary Howard Lutnick on Monday after a powerful AI tool was blocked.”
The Monday meeting with the Department of Commerce is expected to include more documentation of the alleged issue, according to a person familiar with the matter. It remains unclear, however, whether after the meeting Anthropic will be able to make Fable 5 and its more powerful counterpart, Mythos 5, accessible again.
Both tools represent new versions of Claude Mythos, an AI program that caused serious concerns among technology, finance and government leaders when it was first previewed privately in April. At that time, Anthropic said the tool was so intelligent it could be dangerous because of its ability to exploit or hack computer systems. Canadian Finance Minister François-Philippe Champagne told the BBC in April that the attention was warranted, in part, because "it's the unknown, unknown."
Despite an ongoing lawsuit by Anthropic against the US Department of Defense over how its models can be used, US government agencies have been testing Mythos. Tensions appeared to ease a few weeks ago after what was described as a "productive" meeting with senior White House officials.
Anthropic had released Fable 5 with safeguards and user limitations, but acknowledged that "releasing a model this capable comes with risks. Fable's capabilities exceed those of any model we've ever made generally available." The roughly 150 groups that previously had access to preview Mythos will now have access to Claude Mythos 5, which does not have limitations on cybersecurity or biology, depending on an organization's uses. Those groups have reported finding more than 10,000 critical security flaws in their systems.
Dozens of tech leaders from companies including Nvidia, Zoom and Mercedes-Benz, as well as former security staff from the US government and Google, have urged the US government in an open letter to allow Anthropic to release the models publicly. Anthropic co-founder Jack Clark told BBC Newsnight that AI capabilities were expanding so rapidly that there should be a way for the public to slow the technology's advancement. "You want the option to be able to take your foot off the gas and put your foot on the brake," he said.