White House Chief of Staff to Meet With Anthropic Over Cybersecurity AI Model
White House Chief of Staff Susie Wiles plans to meet with Anthropic CEO Dario Amodei to discuss the company's new Mythos AI model, which the federal government views as potentially significant for national security and economic implications.
A White House official said the administration is engaging with advanced AI labs about their models and software security. Any new technology the federal government uses would require a technical evaluation period.
The meeting signals a shift in tone after months of conflict between the Trump administration and Anthropic, a San Francisco-based company known for building safety guardrails into its AI systems.
Background: Recent Tensions
In February, President Trump said the administration "will not do business with them again" after a contract dispute between Anthropic and the Pentagon. Defense Secretary Pete Hegseth sought to declare the company a supply chain risk-an unprecedented action against a U.S. firm.
Anthropic challenged both moves in federal court, saying it wanted assurance the Pentagon would not use its technology in fully autonomous weapons or domestic surveillance. Hegseth argued the Pentagon should be able to use any technology it deemed lawful.
A federal judge blocked Trump's directive in March, preventing agencies from ceasing use of Anthropic products.
What Mythos Does
Anthropic announced Mythos on April 7, describing it as exceptionally capable at finding and exploiting computer vulnerabilities-surpassing human cybersecurity experts in some cases. The company is limiting access to select customers due to these capabilities.
The United Kingdom's AI Security Institute evaluated the model and found it "a step up" over previous versions. The institute noted that Mythos can exploit systems with weak security and that similar models will likely emerge from other companies.
David Sacks, the White House's AI and crypto czar, said the cybersecurity risks warrant serious attention. "As coding models become more capable, they're more capable at finding bugs, finding vulnerabilities, and stringing together multiple vulnerabilities into exploits," he said.
Industry Response and Timeline
Some industry observers initially questioned whether Anthropic was overstating Mythos's risks as a marketing tactic. Sacks said this case appears "more on the real side."
Anthropic co-founder Jack Clark said Mythos, while advanced, is not unique. Other companies will develop similar systems within months, and China will likely release open-weight models with comparable capabilities within a year to 18 months.
Anthropic formed an initiative called Project Glasswing, bringing together Amazon, Apple, Google, Microsoft, JPMorgan Chase, and other organizations to secure critical software against risks the model poses.
The company is also in talks with the European Union about its AI models, including advanced versions not yet released in Europe.
For professionals in AI for Government and AI for IT & Development, these developments highlight how federal agencies are evaluating advanced AI capabilities and the security considerations shaping policy decisions.
Your membership also unlocks:
