Official: White House chief of staff to meet Anthropic CEO regarding dangerous new Mythos model

(SeaPRwire) –   White House Chief of Staff Susie Wiles intends to discuss Anthropic’s new Mythos AI model with the company’s CEO, Dario Amodei—this model has caught the federal government’s eye due to its potential to reshape national security and the economy.

An anonymous White House official (who asked not to be named when talking about Friday’s planned meeting) stated that the administration is collaborating with leading AI laboratories regarding their models and software security. The official emphasized that any new technology the federal government might adopt would require a technical assessment phase.

This meeting follows heightened tensions between the Trump administration and Anthropic—a company focused on AI safety that has worked to establish safeguards for AI development, aiming to reduce potential risks while boosting the technology’s economic and national security advantages for the U.S.

President Donald Trump attempted to bar all federal agencies from using Anthropic’s Claude chatbot amid a contract disagreement between the company and the Pentagon. In a February social media post, Trump declared the administration “will not do business with them again!”

Defense Secretary Pete Hegseth also tried to label Anthropic a supply chain risk—an unprecedented action against a U.S. firm, which Anthropic has contested in two federal courts. The company requested guarantees that the Pentagon wouldn’t use its tech in fully autonomous weapons or the surveillance of U.S. citizens. Hegseth argued the company must permit any Pentagon uses that are legally allowed.

In March, U.S. District Judge Rita Lin handed down a decision that prevented the enforcement of Trump’s social media order, which had directed all federal agencies to cease using Anthropic’s products.

Anthropic chose not to comment on the upcoming meeting.

San Francisco-based Anthropic announced its new Mythos model on April 7, describing it as “strikingly capable.” Due to its ability to outperform human cybersecurity experts in identifying and exploiting computer vulnerabilities, the company is restricting access to the model to a small group of chosen customers.

While some industry experts have doubts about whether Anthropic’s claims of an overly powerful AI are just a marketing tactic, even some of the company’s harshest critics have noted that Mythos could mark a significant step forward in AI development.

David Sacks—an influential critic of Anthropic and former White House AI and crypto czar—stated that people should “take this seriously.”

“Whenever Anthropic is causing alarm, you have to wonder: Is this a strategy? Is it part of their Chicken Little act? Or is it genuine?” Sacks said on the “All-In” podcast, which he co-hosts with other tech investors. “When it comes to cybersecurity, I actually think they deserve credit here—this seems more real than not.”

Sacks explained: “It’s logical that as coding models grow more advanced, they get better at finding bugs. That translates to better ability to spot vulnerabilities, which then means they can connect multiple vulnerabilities to create an exploit.”

Beyond the U.S., the Mythos model’s potential benefits and risks have also drawn notice.

The UK’s AI Security Institute reported that it assessed the new model and found it a “step up” from earlier versions, which were already evolving quickly.

“Mythos Preview can take advantage of systems with poor security, and it’s probable that more models with similar capabilities will be created,” the institute noted in a report.

European Commission spokesman Thomas Regnier stated Friday that Anthropic has been discussing its AI models—including advanced ones not yet launched in Europe—with the European Union.

Axios was the first to report on the planned meeting between Wiles and Amodei.

Alongside the Mythos announcement, Anthropic revealed it’s launching Project Glasswing—an initiative that brings together tech leaders like Amazon, Apple, Google, and Microsoft, plus other firms such as JPMorgan Chase. The goal is to protect the world’s critical software from the “severe” risks the new model could pose to public safety, national security, and the economy.

“We’re making Mythos available to a select group of the world’s key companies and organizations so they can use it to identify vulnerabilities,” Jack Clark—Anthropic co-founder and policy chief—said at this week’s Semafor World Economy conference.

Clark further noted that although Mythos is ahead of its time, it isn’t a “special model.”

“In a few months, other companies will release similar systems, and in one to one-and-a-half years, China will have open-weight models with these capabilities,” he stated. “The world needs to prepare for more powerful systems that will soon be part of our reality.”

___

O’Brien reported from Providence, Rhode Island. AP business reporter Kelvin Chan provided additional reporting from London.