JD Vance was visibly rattled.
The innovation he had fervently supported for months was now posing a serious threat to America’s banks, hospitals, and water systems.
In an urgent call with the nation’s top tech leaders this past April, the Vice President raised the alarm about the potential havoc their creations could wreak.
On the other end of the line were some of the wealthiest individuals globally: Elon Musk, Sam Altman from OpenAI, Dario Amodei of Anthropic, Google’s Sundar Pichai, and Satya Nadella from Microsoft.
“We need to collaborate on this,” Vance reportedly urged them.
With his strong connections to Silicon Valley, Vance was the White House’s key figure on AI issues. Now, he was delivering a sobering message to his tech allies: it was time to rein in their technology.
It marks a stunning about-face.
For months, Donald Trump and his top officials insisted AI restrictions had to be torn down so the US could beat China in the Manhattan Project-style race for technological dominance.
On the April phone call, the Vice President warned that powerful new artificial intelligence models such as Anthropic’s Mythos could wreak chaos on vulnerable small-town hospitals and banks, exposing them to cyberattacks that local governments would be ill-equipped to repel
Some of the richest men in the world were present for the call, including Elon Musk
The catalyst was Mythos, the latest AI model from Silicon Valley giant Anthropic, and its astonishing ability to autonomously hunt down and exploit cybersecurity flaws.
So dangerous is the model that Anthropic has withheld it from public release, fearing Mythos could be turned loose on America’s critical infrastructure.
In testing, it behaved like an elite hacker, uncovering a 27-year-old flaw in OpenBSD, an operating system prized for its security, that had eluded human researchers for decades.
It was also able to discover weaknesses in the Linux kernel, the backbone of most of the world’s servers.
Treasury Secretary Scott Bessent last month summoned the chiefs of America’s biggest banks to Washington for an emergency closed-door meeting on the threat.
The summit was attended by CEOs including Citigroup’s Jane Fraser, Morgan Stanley’s Ted Pick, Bank of America’s Brian Moynihan, Wells Fargo’s Charlie Scharf and Goldman Sachs’s David Solomon.
Vance’s alarm was triggered by a White House briefing on Mythos days before his April call to the tech bosses.
The White House is now weighing an executive order that would formally create an oversight process for the most advanced models, the Wall Street Journal reported.
Mythos is so capable of autonomously hunting down and exploiting cybersecurity flaws that the company responsible for it withheld the program from public release, fearing the model could be used to damage America’s critical infrastructure
Trump officials have asked Anthropic to halt any expansion of Mythos access to additional companies and organizations overseeing critical digital infrastructure.
The Silicon Valley behemoth has restricted Mythos to roughly 40 elite firms under a confidential program dubbed ‘Project Glasswing’, among them Apple, Microsoft, Google, JPMorgan, Goldman Sachs, Bank of America, Citi and Morgan Stanley.
The AI company had deemed the model too dangerous for wide release, but reports indicate it has leaked anyway.
Unauthorized users from a ‘private online forum’ managed to break into Mythos through a third-party vendor in the same week as Vance’s call, Bloomberg reported.
The group reportedly got in by making ‘an educated guess about the model’s online location based on knowledge about the format Anthropic has used for other models.’
The users told Bloomberg they were simply ‘interested in playing around’ with new AI models, not in wreaking havoc.
‘We’re investigating a report claiming unauthorized access to Claude Mythos Preview through one of our third-party vendor environments,’ an Anthropic spokesperson said.
Musk and Peter Thiel, two of the biggest donors to Trump’s 2024 campaign, both hold sizable financial stakes in AI companies, complicating any crackdown.
The relentless rise of Chinese AI, including state-backed labs, adds increased pressure and raises the grim specter of a national security crisis should US companies find themselves restricted by lawmakers in Washington.
Trump flies to China next week to meet Xi Jinping, where AI will loom large over the summit amid tensions over Taiwan – the chip-making hub which Beijing has threatened to invade.
Anthropic has already found itself at the center of America’s defense, securing a two-year Pentagon deal worth up to $200 million to supply AI tools for military and intelligence work.
It leaves the White House caught between two fears: move too fast and AI could trigger the catastrophe Vance warned about, but move too slow and Beijing will leave America in the dust.
