Summary: Anthropic CEO Dario Amodei is in talks with White House Chief of Staff Susie Wiles on Friday to access Mythos, a frontier artificial intelligence model that can identify and exploit thousands of zero-day vulnerabilities in every major operating system and browser. The meeting follows the blacklisting of Anthropic by the Pentagon after Amodei refused to lift security restrictions, and the US Treasury, the intelligence community, CISA and UK financial regulators are seeking access to the model through the Project Glasswing program controlled by Anthropic.
Anthropic CEO Dario Amodei is scheduled Meet White House Chief of Staff Susie Wiles on Friday It represents the most significant step yet to resolve a standoff with the Pentagon over the company’s refusal to remove security restrictions from its AI models. The meeting comes as multiple U.S. government agencies, including the Treasury Department, the intelligence community and the Cybersecurity and Infrastructure Security Agency, seek access to Anthropic’s Mythos model, the frontier artificial intelligence system whose cybersecurity capabilities have prompted emergency briefings from Washington to London to Ottawa.
Announced on April 7, Mythos is not a cybersecurity product. It’s a general-purpose AI model that can identify and exploit thousands of previously unknown zero-day vulnerabilities in every major operating system and web browser during testing. It discovered flaws that survived decades of human security checks and millions of automated tests. When an employee was directed to develop exploits, he succeeded in more than 83% of the first attempts. It is the first AI model to complete a 32-step enterprise network attack simulation from start to finish.
Anthropic chose not to release Mythos to the public. Instead, he created Project Glasswing, a managed access program that provides models to about 40 vetted organizations, including Amazon Web Services, Apple, Google, Microsoft, Nvidia, Cisco, CrowdStrike, JPMorgan Chase and Palo Alto Networks. The company has committed up to $100 million in Mythos usage credits and $4 million in donations to open source security organizations.
Pentagon conflict
The White House meeting is the product of a heated debate since February. Defense Secretary Pete Hegseth has demanded that Anthropic give the Pentagon unfettered access to its models for all legitimate purposes, including potential use in autonomous weapons systems and internal surveillance. Amodei refused. Hegseth designated Anthropic a national security supply chain risk, a label previously reserved for companies linked to foreign adversaries, and blacklisted it from government contracts.
Anthropic sued the Trump administration in early March, filing two federal lawsuits alleging wrongful retaliation. A federal judge initially blocked the blacklist, but an appeals court overturned that decision on April 8, keeping Anthropic out of Department of Defense contracts while litigation continues. The company may still work with other government agencies.
The paradox is that the same government that blacklisted Anthropic now wants access to its most powerful model. The Treasury Department is looking to Mythos to hunt for vulnerabilities in its systems. Parts of the intelligence community and CISA are already testing it. The White House Office of Management and Budget is building safeguards to allow federal agencies to use the controlled version. Axios reported that Anthropic had hired Trumpworld advisers to facilitate the talks, and Friday’s meeting was set to pave the way for a deal.
Why Mythos is important
Jamie Dimon, CEO of JPMorgan Chase, publicly stated that Mythos “discovers more vulnerabilities” to cyberattacks. The UK’s AI Security Institute evaluated the preview version and found it to be “significantly more cyber-attack capable than any model previously evaluated”, noting that it is the first model capable of combining multiple attack steps into full end-to-end intrusions. The Council on Foreign Relations called it a “tipping point for AI and global security.”
For Mythos, the defense case is simple: if an AI model can find vulnerabilities that human security teams and automated testing have missed for decades, giving that model to organizations responsible for protecting critical infrastructure allows them to fix them before adversaries discover them. The risk of an attack is equally simple: the same ability in hostile hands would be disastrous. Anthropic’s decision to restrict access rather than publicly disclose it is a direct application of security principles that put it at odds with the Pentagon.
The the company’s commercial trajectory gives him leverage in negotiations. Anthropic has reached $30 billion in annual revenue, has raised $800 billion in investor proposals and is exploring an IPO. It doesn’t need Pentagon contracts to survive. What he needs is a resolution that preserves his security commitments while restoring his ability to work with the broader US government, something the Wiles meeting is designed to explore.
Global response
The myth has become a topic of concern outside of Washington. Andrew Bailey, the governor of the Bank of England, clearly called it a cybersecurity risk in a speech at Columbia University on April 15. The Bank’s Cross Market Operational Resilience Team is holding emergency briefings over two weeks with the CEOs of eight of the UK’s largest banks, four financial infrastructure providers, two insurers and representatives from the Treasury, the Financial Conduct Authority and the National Cyber Security Centre.
Anthropic plans to provide Mythos access to select British banks during the day As part of Project Glasswing’s expansion and quadrupling its London office to 800 staff in King’s Cross. The UK’s AI Security Institute, which has an existing evaluation partnership with Anthropic, published its technical evaluation on April 17. Canadian Finance Minister François-Philippe Champagne described Mythos as an “unknown unknown” discussed at IMF meetings. Global regulators are coordinating on how to assess and manage the consequences of cyber security.
The geopolitical dimension it is inevitable. The US government’s request for access to Mythos is in tension with the punishment of the company that built it. Anthropic’s willingness to provide the model to British banks and regulators amid a lawsuit with the Pentagon creates a situation where America’s closest ally could gain access to a critical national security tool before its own government. This dynamic gives the White House an incentive to resolve the dispute, which goes beyond the initial disagreement over security bars.
What a contract might look like
An outline of a potential solution appears. Anthropic will restore eligibility for government contracts and provide access to Mythos for defense cybersecurity purposes. Pentagon to withdraw supply chain risk designation. Anthropic will maintain its restrictions on autonomous weapons and mass surveillance applications, but potentially agree to a process for reviewing specific military use cases that do not cross those lines. Both sides have reasons to compromise: Anthropic because the Blacklist hurts him reliability of the enterpriseand management because it needs technology.
Whether Amodei and Wiles reach such a deal on Friday, or simply begin the process of getting there, is less important than what the meeting represents. The company that built the most capable cybersecurity tool available did so as a byproduct of building a general-purpose AI model, then restricted its release for security reasons, then got punished by the government for upholding the same security principles, and is now mocked by that government because the tool is too worth ignoring.
This sequence reflects something important about where AI governance stands in April 2026. Technology is evolving faster than the institutions responsible for managing it can keep up, and companies that take security seriously are both rewarded by the market and punished by the state. The Mythos is the starkest example of a model whose capabilities are so consequential that limiting and releasing are both defensible positions, and the debate between them plays out in the West Wing, not in a research paper or a congressional hearing.
