The Trump administration is set for a pivotal, high-stakes meeting with Anthropic’s top leadership on Friday, marking a major turn in a months-long public and legal conflict between the U.S. government and one of the world’s leading artificial intelligence developers. White House Chief of Staff Susie Wiles will open discussions with Anthropic CEO Dario Amodei, centered on the company’s newly unveiled Mythos model, a cutting-edge AI system that has drawn unprecedented federal scrutiny over its far-reaching implications for both U.S. national security and economic competitiveness.
A senior White House official, speaking on condition of anonymity to detail private planning for the gathering, confirmed that the current administration is proactively engaging with leading advanced AI research labs to review new model capabilities and assess software security protocols. The official emphasized that any AI technology under consideration for future federal government use would first undergo a rigorous, extended technical evaluation process to verify safety and functionality before any official adoption.
This planned meeting comes after months of escalating friction between the Trump administration and Anthropic, a San Francisco-based AI firm that has long prioritized building guardrails around advanced AI development to mitigate catastrophic risk while advancing potential benefits for the U.S. The dispute erupted earlier this year when President Donald Trump issued a public social media order banning all federal agencies from using Anthropic’s flagship chatbot Claude amid a bitter contract conflict with the Pentagon. In the February post, Trump declared the administration “will not do business with them again!”
Defense Secretary Pete Hegseth doubled down on the administration’s pressure, pushing to designate Anthropic as a federal supply chain risk—an unprecedented move against a domestic U.S. technology company. Anthropic has challenged that designation in two separate federal courts. The core of the dispute centers on Anthropic’s demand for binding assurances that the Pentagon will not use its AI technology to develop fully autonomous weapons or conduct mass surveillance of U.S. civilians, while Hegseth has insisted the company must permit all lawful uses the Pentagon deems appropriate. In a landmark March ruling, U.S. District Judge Rita Lin blocked enforcement of Trump’s original ban on federal use of Anthropic products, handing the company a major legal victory.
What has reignited cross-government and global attention on Anthropic in recent weeks is the April 7 launch of Mythos, a model the company describes as “strikingly capable” of outperforming professional human cybersecurity experts at identifying and exploiting critical software vulnerabilities. Because of this unprecedented capability, Anthropic has restricted access to Mythos, only rolling it out to a small, curated group of vetted customers.
While some tech industry analysts have questioned whether Anthropic’s warnings about Mythos’ power amount to a calculated marketing tactic, even prominent critics of the company have acknowledged the model likely represents a meaningful leap forward in AI capability. David Sacks, the White House’s own AI and crypto czar and a frequent Anthropic critic, told listeners of his popular “All-In” podcast that the claims around Mythos should be taken seriously. “Anytime Anthropic is scaring people, you have to ask, ‘Is this a tactic? Is this part of their Chicken Little routine? Or is it real?’” Sacks said. “With cyber, I actually would give them credit in this case and say this is more on the real side.” He added that the logic of advancing AI capability holds: as coding models grow more powerful, they naturally gain improved ability to find security bugs, chain multiple vulnerabilities together, and develop functional exploits that can compromise protected systems.
The model’s unique combination of transformative benefits and catastrophic risk has drawn attention from global policymakers beyond U.S. borders. The United Kingdom’s AI Security Institute, which conducted its own independent evaluation of Mythos, concluded the model is a clear “step up” from already rapidly improving earlier AI generations. The institute warned in its report that “Mythos Preview can exploit systems with weak security posture, and it is likely that more models with these capabilities will be developed” across the global tech sector in the coming years. European Union officials also confirmed Friday that Anthropic has held ongoing talks with EU regulators about Mythos and other advanced, unreleased AI models.
Alongside the launch of Mythos, Anthropic unveiled Project Glasswing, a cross-industry collaborative initiative bringing together tech giants including Amazon, Apple, Google, and Microsoft, plus major financial institutions like JPMorgan Chase, to harden global critical software infrastructure against the risks posed by next-generation AI systems like Mythos. The company says restricted access to Mythos allows key public and private sector organizations to use the model’s own capability to find and patch unaddressed vulnerabilities before bad actors can exploit them.
Speaking at this week’s Semafor World Economy conference, Anthropic co-founder and policy chief Jack Clark stressed that Mythos, while ahead of current industry curves, is not an anomaly. “There will be other systems just like this in a few months from other companies, and in a year to a year-and-a-half later, there will be open-weight models from China that have these capabilities,” Clark said. “So the world is going to have to get ready for more powerful systems that are going to exist within it.”
The meeting between Wiles and Amodei was first reported by Axios. Anthropic declined to comment on the planned gathering ahead of time. Reporting for this article was contributed by AP business reporter Kelvin Chan from London.