JD Vance held an ad-hoc AI safety summit last month, gathering a conference call with the nation’s top tech CEOs after a new Anthropic model showed how easily it could autonomously find, and then exploit, the types of cybersecurity systems that guard against attacks on banks, hospitals, infrastructure and just about anything else.
The April phone call included Elon Musk, OpenAI CEO Sam Altman, Google CEO Sundar Pichai, Microsoft CEO Satya Nadella and Anthropic CEO Dario Amodei, according to multiple media reports. The message from Vance: “We all need to work together on this.”
The vice president’s warning – that the new batch of AI models are now arriving with hacking superpowers – was sparked by one such model, called Mythos, created by Anthropic. Mythos reportedly easily found flaws in venerated security firewalls like OpenBSD and the Linux kernel, which drives many servers across the globe, and has been withheld from public release.
Vance – previously a champion of Trump’s AI deregulation push – became increasingly concerned following a White House briefing on Mythos. According to multiple reports, the White House is considering reversing course with an executive order creating formal oversight for AI’s most advanced systems.
Also last month, Treasury Secretary Scott Bessent reportedly assembled the chief executives from several major U.S. banks to Washington for a closed-door meeting on AI-related cybersecurity threats. In attendance were Morgan Stanley CEO Ted Pick, Bank of America CEO Brian Moynihan, Citigroup CEO Jane Fraser, Wells Fargo CEO Charlie Scharf and Goldman Sachs CEO David Solomon.
The Trump administration has reportedly asked Anthropic to pause expansion of Mythos access, which now spans some 40 companies including Apple, Microsoft, Google, JPMorgan, Goldman Sachs, Bank of America, Citi and Morgan Stanley. Bloomberg reported last month – the same week as Vance’s call – that Mythos had been accessed by unauthorized users who got in through one of those third-party vendors.
