clauding.de
DE
• 2 min read AI-generated

NSA Uses Mythos Despite Pentagon Feud — Anthropic's Most Powerful Model Goes to the Spies

Copy article as Markdown

While the Pentagon labels Anthropic a 'supply-chain risk,' the NSA is quietly using their restricted Mythos model for vulnerability scanning.

Anthropic Mythos NSA Pentagon Cybersecurity Intelligence
Featured image for "NSA Uses Mythos Despite Pentagon Feud — Anthropic's Most Powerful Model Goes to the Spies"

The most absurd story in the AI world just got a new chapter. The NSA is reportedly using Anthropic’s Mythos model — the same model deemed too powerful for public release. That’s according to Axios, confirmed by TechCrunch and CNBC.

The irony in three sentences

The Pentagon designated Anthropic as a “supply-chain risk” because the company refused to give the military unrestricted access to Claude. Meanwhile, the NSA — an agency under the Pentagon’s own umbrella — is using Anthropic’s most powerful and most restricted model. Let that sink in.

What the NSA is doing with Mythos

According to reports, the NSA is using Mythos primarily for scanning environments for exploitable vulnerabilities. This aligns with Anthropic’s own positioning: Mythos was introduced as a cybersecurity model that can find security flaws faster than human teams can patch them.

Anthropic has restricted Mythos access to roughly 40 organizations, publicly naming only about a dozen. Goldman Sachs is on the list, the UK’s AI Security Institute too. And now, apparently, the NSA.

The bigger picture

The relationship between Anthropic and the Trump administration appears to be thawing. Last week, Dario Amodei met with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent — a meeting both sides called “productive.”

That’s a remarkable shift. Just a month ago, Anthropic sued the administration twice. The preliminary injunction against the Pentagon blacklist was initially granted, then overturned on appeal.

My take

This story perfectly captures how absurd the current situation is. One arm of the US defense apparatus labels Anthropic a risk, while another arm uses their cutting-edge technology for national security. That’s not a contradiction — that’s bureaucracy in real time.

For Anthropic, the NSA’s use of Mythos is still a powerful signal. When the world’s most powerful intelligence agency picks your model for cybersecurity work, that’s the ultimate seal of approval. Mythos apparently finds vulnerabilities that nothing else can — and that’s precisely why it’s both so dangerous and so valuable.

Sources: