AI cybersecurity – JD Vance convened major tech leaders over new AI models’ hacking potential, as the White House weighs tougher oversight amid bank security talks.
JD Vance sounded an urgent warning to the tech industry, pressing top executives to confront an uncomfortable new reality: AI systems are arriving with capabilities that can probe and potentially exploit cybersecurity defenses.
The vice president held an ad-hoc AI safety summit last month by organizing a conference call with the nation’s leading technology CEOs after an Anthropic model reportedly demonstrated how easily it could autonomously locate and exploit weaknesses in cybersecurity systems used to protect banks. hospitals. critical infrastructure. and other high-stakes targets.
According to multiple reports. the April call included Elon Musk. OpenAI CEO Sam Altman. Google CEO Sundar Pichai. Microsoft CEO Satya Nadella. and Anthropic CEO Dario Amodei.. The message attributed to Vance was straightforward: the country needs the industry to work together on AI safety and the security risks that come with more powerful systems.
The warning reportedly centered on a model called Mythos.. It was described as being able to find flaws in well-regarded security defenses. including OpenBSD and the Linux kernel—core components that underpin servers across the internet.. The model has also reportedly been withheld from public release, even as internal access has expanded.
Vance’s concern, as multiple reports outlined, intensified after a White House briefing on Mythos.. In that telling. the administration is weighing a shift in policy. with the White House considering reversing course through an executive order that would create formal oversight for the most advanced AI systems.
That possible change comes against a backdrop of prior political positions on AI. Vance was previously described as a champion of President Trump’s push for AI deregulation, making the latest emphasis on oversight and coordinated safety measures notable.
While the call with major tech leaders focused on the broader AI risk picture. the administration’s attention to cybersecurity appeared to reach directly into the banking sector as well.. It was reported that Treasury Secretary Scott Bessent convened senior bank executives in Washington for a closed-door meeting focused on AI-related cybersecurity threats.
Multiple reports said the closed meeting included 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 gathering underscored how the concerns raised by Mythos are not abstract theory. but framed as a practical problem for institutions that rely on complex. layered defenses.
In another reported development, the Trump administration asked Anthropic to pause expansion of Mythos access. Despite that, it was reported that Mythos access now spans about 40 companies, including Apple, Microsoft, Google, JPMorgan, Goldman Sachs, Bank of America, Citi, and Morgan Stanley.
The risk alarm also appears to have been reinforced by an earlier report indicating that Mythos was accessed by unauthorized users.. Bloomberg reported last month—during the same general week as Vance’s call—that access by unauthorized parties occurred through one of the third-party vendors connected to the system.
For executives. the situation puts a spotlight on the full ecosystem around advanced AI systems—not just the model itself. but the network of partners. vendors. and permissions that determine who can reach powerful capabilities.. If third-party pathways can be used to reach restricted systems. then “security” becomes as much an operations and governance challenge as a technical one.
It also raises hard questions about how AI safety efforts translate into day-to-day standards across industries.. When a model can reportedly probe or exploit elements of cybersecurity infrastructure. the implications extend beyond any single company or sector. affecting how banks. hospitals. and public services evaluate risk when they adopt or integrate frontier AI.
Meanwhile. the administration’s reported shift toward formal oversight suggests officials may be looking to create clearer guardrails for the most capable systems—especially as access expands across companies and the attack surface grows.. Vance’s call to “work together” now reads less like a generic plea and more like a push for coordinated action while the policy conversation is still in motion.
JD Vance AI safety Mythos model AI cybersecurity Anthropic executives White House oversight bank cybersecurity threats tech CEO summit