A powerful AI model has already uncovered thousands of critical security vulnerabilities. These include flaws in every major operating system and web browser. Now it sits at the centre of a growing intelligence community debate in Washington.
Anthropic unveiled Project Glasswing on Tuesday, a defensive cybersecurity initiative built around Claude Mythos Preview. The model remains withheld from public release. Partners with controlled access include Amazon Web Services, Apple, Cisco, Google and Microsoft.
Crucially, the scale of what Mythos Preview has already found is what is alarming officials. Security researchers had long anticipated a model capable of identifying vulnerabilities at speed. That moment has now arrived.
Intelligence Agencies Are Taking Notice
Multiple US intelligence agencies and Defence Department components are already discussing the implications. Sources familiar with internal deliberations confirmed this. Anthropic says it has briefed senior officials at two key bodies. These are the Cybersecurity and Infrastructure Security Agency and NIST’s Centre for AI Standards and Innovation.
Meanwhile, NSA analysts have been discussing the release informally. The concern is simple. A tool powerful enough to expose weaknesses in critical infrastructure could be devastating in the hands of an adversary.
“How is anyone supposed to defend against all of this at once?” said one executive at a cyber investment firm.
Hayden Smith, co-founder of software supply chain risk firm Hunted Labs, was more direct. Given ongoing attacks on the open-source ecosystem, he argued that Mythos reaching a hostile government was “barely a hypothetical.”
Offence and Defence Are the Same Problem
Morgan Adamski, former executive director at US Cyber Command, framed the core dilemma clearly. Finding a vulnerability in an adversary’s network raises an immediate question: does the same flaw exist in US systems?
“There’s going to be a real equity conversation that occurs,” she said.
Senator Mark Warner, vice chairman of the Senate Intelligence Committee, called on industry to accelerate patching. As AI speeds up vulnerability discovery, he warned, patching must keep pace.
The announcement also arrives against a fraught backdrop. Earlier this year, the Pentagon designated Anthropic a supply chain risk. The reason: the company had declined to ease restrictions on domestic surveillance tools and autonomous weapons for Pentagon use. As a result, the White House ordered federal agencies to phase out Anthropic tools. Anthropic is contesting that decision in court.
Click Here For The Original Source.
