What Is Mythos?
Anthropic's Claude Mythos Preview is its most powerful AI model yet capable of finding and exploiting software vulnerabilities autonomously. Here is what it is, what it can do, and why Anthropic is restricting its release to 40 organizations through Project Glasswing.
Published: 4/11/2026, 06:52:41 UTC | Updated: 4/11/2026, 07:20:02 UTC
What Is Mythos? Anthropic's Most Powerful AI Model - And Why It Won't Be Released to the Public
When Anthropic named its newest AI model "Mythos," it was signaling something the company believed was genuinely unprecedented - a model so capable at finding and exploiting software vulnerabilities that releasing it publicly posed risks it was not prepared to accept.
What is Claude Mythos Preview?
Claude Mythos Preview is Anthropic's latest and most powerful AI model, announced on April 7, 2026. It is a general-purpose model - not specifically built for cybersecurity - but its advanced coding and reasoning capabilities produced an extraordinary ability to identify software vulnerabilities that had gone undetected for years, in some cases decades.
The announcement came alongside a new initiative called Project Glasswing - a coordinated effort to deploy Mythos's capabilities defensively, giving major technology companies a head start in patching critical vulnerabilities before such capabilities become more widely available.
What can it do?
During internal testing, Mythos Preview identified thousands of previously unknown zero-day vulnerabilities across every major operating system and every major web browser. Many had survived years of human security review and millions of automated tests. The oldest vulnerability it identified was a 27-year-old flaw in OpenBSD - an operating system built with security as its primary design principle.
Beyond finding vulnerabilities, Mythos can exploit them - autonomously constructing working attack code without human direction. In one documented case, it wrote a browser exploit that chained together four separate vulnerabilities to escape both renderer and operating system sandboxes.
Why is Anthropic not releasing it publicly?
The same capability that makes Mythos valuable to defenders makes it dangerous in the wrong hands. A model that can find and exploit critical vulnerabilities in major software systems - autonomously, at scale - significantly lowers the barrier to sophisticated cyberattacks.
Anthropic made a deliberate decision to limit access rather than release broadly. This marks the first time a leading AI company has publicly withheld a model over safety concerns since OpenAI restricted GPT-2 in 2019.
Who has access?
Access is limited to approximately 40 organizations through Project Glasswing. Launch partners include Amazon, Apple, Microsoft, Nvidia, Broadcom, Cisco, CrowdStrike, Palo Alto Networks, and the Linux Foundation. Anthropic has committed up to $100 million in usage credits to support the initiative. These organizations will use Mythos to scan and secure both their own software and critical open-source infrastructure - then share what they learn with the broader industry.
What is Project Glasswing?
Project Glasswing takes its name from the glasswing butterfly - chosen because its transparent wings reflect how software vulnerabilities are present and potentially dangerous, yet difficult to see. The initiative is Anthropic's attempt to put Mythos's capabilities to work for defense before they proliferate more broadly.
Why does it matter for banks and financial institutions?
Federal Reserve Chair Jerome Powell and Treasury Secretary Scott Bessent convened a special gathering with major US bank CEOs this week to discuss the cyber risks posed by advanced AI models, including Mythos. The meeting reflected growing concern among regulators about the implications of AI systems capable of identifying and exploiting software vulnerabilities at scale - risks that financial infrastructure, which runs entirely on software, cannot ignore.
What comes next?
Anthropic has stated that Mythos Preview is not the end point - it is a preview of a class of models whose capabilities will continue to advance. The company's goal through Project Glasswing is to ensure that defenders gain experience with these capabilities before they become widely available.
OpenAI is reportedly developing a model internally known as "Spud" with comparable cybersecurity capabilities - suggesting that what Mythos represents today will become an industry-wide reality within months, not years.
The window for defenders to get ahead is open. Project Glasswing is Anthropic's attempt to use it.