Anthropic Mythos AI Cybersecurity Threat Brings Amodei Back to the White House

When we covered Project Glasswing earlier this month, the story was about a model too dangerous to release publicly and what Anthropic decided to do with it instead. That story has moved.

On Friday, Anthropic CEO Dario Amodei walked into the West Wing for a meeting with White House Chief of Staff Susie Wiles. Treasury Secretary Scott Bessent was also in the room. The White House called the talks “productive and constructive.” Anthropic said the same. When a reporter asked President Trump about the visit on a runway in Phoenix, he responded “Who?” and said he had “no idea” Amodei was there.

That detail aside, the meeting itself is one of the more striking political reversals in recent AI history. Just weeks ago, the Trump administration had declared Anthropic a supply chain risk – a designation ordinarily reserved for foreign adversaries – and Trump himself said the administration would “not do business with them again.”

A federal judge in San Francisco has since blocked the enforcement of that directive, keeping Anthropic eligible to work with non-military agencies while the litigation plays out. The Pentagon dispute remains very much alive.

What changed the calculus – at least at the White House level – was Anthropic Mythos AI cybersecurity ability. Specifically, the fact that agencies are purportedly watching Mythos do things no other tool can, and are not willing to sit that out.

The model and the politics

As we reported when Anthropic unveiled Project Glasswing, Mythos Preview was not trained specifically for security work. Its ability to autonomously identify and exploit software vulnerabilities emerged from general improvements in reasoning and code, and what it has found since deployment has been striking.

During internal testing, Mythos located thousands of previously unknown, high-severity vulnerabilities in every major operating system and web browser, including a 27-year-old bug in OpenBSD and a 16-year-old flaw in FFmpeg that had passed automated testing five million times without detection.

Rather than ship it publicly, Anthropic released it only to a select group of organisations through Project Glasswing – a coalition that includes AWS, Apple, Cisco, Google, Microsoft, Nvidia, CrowdStrike, and JPMorganChase, among others – backed by up to US$100 million in use credits. The model is being used offensively, in a controlled sense: finding the vulnerabilities before someone else does.

The US government has been watching that coalition operate and wanting in. Intelligence agencies and the Cybersecurity and Infrastructure Security Agency are already testing Mythos, and the Treasury Department has also expressed interest, according to Axios. Treasury and other government agencies have expressed interest in joining the Glasswing list, and before Friday’s White House meeting, two sources told Axios a deal along those lines could be struck soon.

In a separate Axios report, a concern brought up is that Mythos and other cutting-edge AI tools could allow hackers to breach the US financial system. Alternatively, the report reckoned companies and government agencies could use Mythos to harden their cyber defences before bad actors get access.

That dual-use tension is now squarely a political problem. National Cyber Director Sean Cairncross is set to lead a group of federal officials to identify security vulnerabilities in critical infrastructure and strengthen government systems against AI exploitation.

Where the standoff stands

The Friday meeting was engineered to separate two conversations that had become entangled. Going into the session, both sides sought to wall off the Pentagon fight from how the rest of the government engages with Anthropic and next steps are expected to be about how other departments access Mythos Preview, per sources familiar with the negotiations.

One Trump adviser told Axios: “This is a big problem. Everyone’s complaining. There’s all this drama. So this got elevated to Susie to hear Dario out, determine what is bull and start to plot a way forward.”

An administration official summarised the current dynamic succinctly: “There’s progress with the White House. There’s no progress with [the Department of] War.” That split is telling. Civilian agencies like the Departments of Energy and Treasury are responsible for safeguarding critical sectors, like the electric grid and the financial system.

Their concerns are not about autonomous weapons or surveillance. They want the ability Mythos offers, and they are not willing to be collateral damage in a fight between the Pentagon and an AI company. The DOD has not commented on Mythos but has continued using Anthropic’s Claude models in the war with Iran. That footnote is worth sitting with.

Publicly, Anthropic has also been making moves that signal it understands how Washington works. Public filings show Anthropic recently hired lobbying firm Ballard Partners – where Wiles worked for years – specifically for advocacy regarding Department of War procurement.

What comes next

The litigation has not ended. A federal appeals court denied Anthropic’s request to temporarily block the Pentagon’s blacklisting; a San Francisco judge granted a preliminary injunction in a separate case. Anthropic remains barred from DoD contracts but can continue working with the rest of the government while both cases run their course.

The White House said it plans to continue dialogue with Anthropic and other AI companies, and the Office of Management and Budget is already preparing to give agencies access to Mythos to assess their defences, according to Bloomberg. That is meaningful progress, even if the Pentagon remains the unresolved piece.

One source close to the negotiations put it plainly: “It would be grossly irresponsible for the US government to deprive itself of the technological leaps that the new model presents. It would be a gift to China.”

That framing – less about Anthropic’s legal standing, more about what the US cannot afford to give up – is what brought Amodei into the West Wing. Whether the Pentagon ever follows is a different question.

See also: Anthropic’s refusal to arm AI is exactly why the UK wants it

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. This comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.