Inside the Talks Between Anthropic CEO and the White House — Dario Amodei Meets Susie Wiles Over Mythos and the US AI Race
AI / AI Tech Trends | 4 min read
Anthropic CEO Dario Amodei met with White House Chief of Staff Susie Wiles and other senior administration officials on 17 April 2026, in what both sides described as a "productive and constructive" discussion — the most significant engagement between the two parties since a bitter dispute with the Pentagon led to Anthropic being designated a "supply chain risk" and the company filing suit against the federal government. Treasury Secretary Scott Bessent was also present, according to sources. The meeting, first reported by Axios, addressed three areas: cybersecurity, America's lead in the AI race, and AI safety — with Anthropic's newly announced Mythos model, a highly capable AI system with advanced offensive and defensive cybersecurity capabilities, central to the discussions. Anthropic said the meeting "reflected Anthropic's ongoing commitment to engaging with the US government on the development of responsible AI" and that it was "looking forward to continuing these discussions."
Mythos — The AI Model at the Centre of the Negotiations
Claude Mythos — Anthropic's powerful next-generation AI model announced earlier in April 2026 — has been described by the company and independent experts as capable of identifying thousands of vulnerabilities in commonly-used software and constructing sophisticated exploit chains from them. Anthropic has been in "ongoing discussions" with US government officials about Mythos' offensive and defensive cyber capabilities. The company has framed the model's emergence as a reason why the US and its allies must maintain a decisive lead in AI technology, and has said it is "ready to work with local, state and federal representatives to assist in these tasks." Anthropic is deploying Mythos through Project Glasswing — a new cybersecurity initiative that brings together major technology companies including Amazon, Apple, Google, and Microsoft, along with financial institutions including JPMorgan Chase — releasing it to a select group of the world's most important companies and organisations so they can use it to find and address vulnerabilities in their systems. Anthropic co-founder and policy chief Jack Clark explained at the Semafor World Economy conference that the company is releasing Mythos "to a subset of some of the world's most important companies and organizations so they can use this to find vulnerabilities." CISA (the Cybersecurity and Infrastructure Security Agency) and segments of the US intelligence community are evaluating the technology, and the Office of Management and Budget has been preparing agencies to access it.
The Backdrop — Pentagon Dispute, Legal Battles, and a Potential Thaw
The meeting took place against a complex backdrop of legal and political tensions. The dispute between Anthropic and the Pentagon began when the Department of Defense sought unfettered access to Claude for "all lawful purposes" — including use in autonomous weapons systems and for broad surveillance — while Anthropic sought assurances that the technology would not be used in those contexts, arguing that AI models are not yet reliable enough for fully autonomous lethal applications and that US law has not caught up to protect Americans around AI-driven mass surveillance. After negotiations broke down, Defence Secretary Pete Hegseth declared Anthropic a "supply chain risk" — an unprecedented designation for a US company — effectively blacklisting it from the government. Anthropic filed suit in two federal courts. US District Judge Rita Lin blocked enforcement of a presidential directive ordering federal agencies to cease use of Anthropic products, though the DC Circuit Court of Appeals separately ruled it would not bar the DOD from cutting ties with Anthropic while legal proceedings continue. Notably, the DOD has continued to use Anthropic's Claude models in the ongoing conflict. Parallel to the White House engagement, Anthropic confirmed it has also been in talks with the European Union about its AI models, including advanced models not yet released in Europe, confirmed by European Commission spokesman Thomas Regnier on the same day.
Key Takeaways
- • Anthropic CEO Dario Amodei met with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent on 17 April 2026 in discussions described by both sides as "productive and constructive." The meeting — first reported by Axios — addressed cybersecurity, America's lead in the AI race, and AI safety. Anthropic said it is "looking forward to continuing these discussions." The meeting is widely seen as a potential thaw in a significant political and legal standoff between Anthropic and the Trump administration.
- • Claude Mythos — the AI model at the centre of discussions: Anthropic's next-generation AI system capable of identifying thousands of vulnerabilities in commonly-used software and constructing sophisticated exploit chains. Described as a potential "watershed" moment for cybersecurity by Anthropic and independent experts. Being deployed through Project Glasswing — a new cybersecurity initiative with Amazon, Apple, Google, Microsoft, and JPMorgan Chase — released to a select group of major companies to identify and address software vulnerabilities. CISA and US intelligence community are evaluating it; Office of Management and Budget preparing agencies for access.
- • Anthropic's position: Anthropic co-founder and policy chief Jack Clark has stated the company is releasing Mythos to a subset of the world's most important companies to find vulnerabilities — not publicly. Anthropic's blog states: "securing critical infrastructure is a top national security priority for democratic countries — the emergence of these cyber capabilities is another reason why the US and its allies must maintain a decisive lead in AI technology." The company says it is ready to work with government representatives on assessing and mitigating national security risks associated with AI.
- • The dispute background: the Pentagon sought unfettered Claude access for "all lawful purposes" including autonomous weapons and mass surveillance; Anthropic sought assurances against those uses. After negotiations broke down, Defence Secretary Hegseth declared Anthropic a "supply chain risk" — unprecedented for a US company — effectively blacklisting it. Anthropic sued in two federal courts. Judge Rita Lin blocked enforcement of a presidential directive ordering agencies to cease using Anthropic products. The DC Circuit Court separately ruled it will not bar DOD from cutting ties while litigation continues. DOD has continued to use Claude in the ongoing conflict.
- • Broader context: earlier in April, VP JD Vance and Treasury Secretary Bessent met with Amodei, OpenAI's Sam Altman, and other top tech executives on AI cybersecurity. Anthropic has also hired Ballard Partners — a lobbying firm with strong ties to the Trump administration — since the Pentagon designation. On the same day as the White House meeting, the European Commission confirmed Anthropic is also in talks with the EU about its AI models, including advanced unreleased models. The meeting is the second high-stakes engagement Amodei has had with top Trump administration officials in 2026, following a February meeting with Defence Secretary Hegseth.