Anthropic Deploys Advanced AI Model After Security Review

# Nvidia # Google # Microsoft # Anthropic # agentic ai # AI cybersecurity # ai security # amazon web services # Claude Mythos # CrowdStrike # cybersecurity # open source security # Palo Alto Networks # Project Glasswing # vulnerability # zero-day

Anthropic's most advanced AI model has already uncovered thousands of cybersecurity vulnerabilities across every major operating system and web browser. Instead of releasing it publicly, the company discreetly provided it to the organizations tasked with safeguarding the internet's core infrastructure.

That model is Claude Mythos Preview, and the initiative is named Project Glasswing.

Initial partners include Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, the Linux Foundation, Microsoft, Nvidia, and Palo Alto Networks.

Beyond this core group, Anthropic has granted access to over 40 additional organizations that develop or maintain critical software infrastructure. The company is committing up to $100 million in usage credits for Mythos Preview across the initiative, alongside $4 million in direct donations to open-source security organizations.

A Model That Surpassed Its Own Benchmarks

Mythos Preview was not specifically trained for cybersecurity. Anthropic stated that its capabilities "emerged as a natural consequence of broad improvements in code generation, reasoning, and autonomous operation." The same advancements that make the model adept at fixing vulnerabilities also enhance its ability to discover and exploit them.

This dual capability is significant. Mythos Preview has advanced to the point where it largely maxes out existing security benchmarks, compelling Anthropic to focus on novel, real-world challenges—specifically, zero-day vulnerabilities. These are security flaws previously unknown to the software's own developers.

Among its discoveries is a 27-year-old bug in OpenBSD, an operating system renowned for its strong security. In another instance, the model autonomously identified and exploited a 17-year-old remote code execution vulnerability in FreeBSD—CVE-2026-4747. This flaw allows any unauthenticated user on the internet to gain complete control of a server running the NFS service. No human intervention was required beyond the initial instruction to find a bug.

Nicholas Carlini from Anthropic's research team highlighted the model's skill in chaining vulnerabilities together: "This model can construct exploits from three, four, or sometimes five distinct vulnerabilities that, in sequence, achieve a highly sophisticated outcome. I've discovered more bugs in the last few weeks than I have in my entire career prior to this."

Why Isn't It Being Released Publicly?

"We do not plan to make Claude Mythos Preview generally available due to its advanced cybersecurity capabilities," said Newton Cheng, Frontier Red Team Cyber Lead at Anthropic. "Given the rapid pace of AI progress, capabilities like these will likely proliferate soon, potentially reaching actors who may not deploy them safely. The potential consequences for economies, public safety, and national security could be severe."

This concern is not theoretical. Anthropic previously disclosed what it called the first documented case of a cyberattack primarily executed by AI—a Chinese state-sponsored group used AI agents to autonomously infiltrate roughly 30 global targets, with AI handling most tactical operations independently.

The company has also privately briefed senior U.S. government officials on Mythos Preview's full capabilities. The intelligence community is now actively assessing how such a model could transform both offensive and defensive cyber operations.

Addressing the Open-Source Challenge

A key aspect of Project Glasswing extends beyond its headline partners: securing open-source software. Jim Zemlin, CEO of the Linux Foundation, stated plainly: "Historically, security expertise has been a luxury for organizations with large security teams. Open-source maintainers, whose software underpins much of the world's critical infrastructure, have often been left to manage security on their own."

To address this, Anthropic has donated $2.5 million to Alpha-Omega and OpenSSF through the Linux Foundation, and $1.5 million to the Apache Software Foundation. This provides maintainers of critical open-source projects access to AI-powered vulnerability scanning at a scale previously unavailable to them.

What Happens Next

Anthropic states its long-term goal is to deploy Mythos-class models at scale, but only after implementing new safeguards. The company plans to introduce these safeguards first with an upcoming Claude Opus model, allowing for refinement with a less risky system before applying them to Mythos Preview.

The competitive landscape is already shifting. When OpenAI released GPT-5.3-Codex in February, it labeled the model as the first it classified as "high-capability" for cybersecurity under its Preparedness Framework. Anthropic's Glasswing initiative signals that leading AI labs view controlled, limited deployment—not open release—as the emerging standard for models with this level of capability.

Whether this standard will hold as these powerful capabilities become more widespread remains an open question that no single initiative can definitively answer.

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 the AI & Big Data Expo happening 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.

Related article

Comments (0)