BeyondTrust Enters AI Arms Race with Anthropic to Secure Critical Systems

ATLANTA, GA – June 08, 2026 – In a move that signals a significant escalation in the cyber defense of essential services, identity security leader BeyondTrust announced today it has joined Project Glasswing, an elite collaborative initiative by AI safety pioneer Anthropic. The partnership grants BeyondTrust access to Claude Mythos Preview, a frontier AI model, to proactively discover and remediate vulnerabilities across its software portfolio, which is integral to protecting digital infrastructure worldwide.

BeyondTrust’s inclusion in an expanded cohort of Project Glasswing members places it alongside other vendors whose code is foundational to global commerce, government, and healthcare. The initiative marks a pivotal moment in cybersecurity, where the very AI technologies that accelerate threats are now being harnessed for a new, powerful form of collective defense.

The New Frontline: AI vs. AI in Cybersecurity

The cybersecurity landscape is undergoing a tectonic shift, evolving into what many experts describe as an “AI arms race.” As attackers begin to leverage artificial intelligence to discover and weaponize software flaws at machine speed, the window between vulnerability discovery and exploitation has collapsed from months to mere seconds. Project Glasswing is Anthropic’s direct response to this reality.

The project provides partners with access to Claude Mythos Preview, an unreleased AI model so advanced in its ability to find security flaws that Anthropic has restricted its public availability over safety concerns. Since its launch in early April, the model has already demonstrated its staggering potential, helping program partners uncover more than 10,000 high or critical-severity vulnerabilities. In one notable case, Mythos identified a 27-year-old bug in the OpenBSD operating system, a discovery that eluded human experts for decades.

For BeyondTrust, whose software governs privileged access for millions of workloads, joining this initiative is a strategic imperative. “As AI reshapes both software development and cyber defense, the organizations responsible for securing critical infrastructure must continuously raise the bar,” said Janine Seebeck, Chief Executive Officer at BeyondTrust. “Project Glasswing gives us an opportunity to do exactly that by continuing to strengthen the security and resilience of the software our customers rely on to protect all human, machine, and agentic identities.”

Taming the Wild West of AI Identities

While the headline capability of Mythos is finding flaws in code, the partnership’s deeper significance lies in addressing a new and rapidly expanding threat vector: agentic AI identities. These are autonomous AI agents—non-human entities operating with high levels of privilege to access sensitive data and execute critical business processes. They represent what the press release calls one of the “fastest-growing yet least-governed classes of privileged identities.”

This is the new frontier of identity security, a digital “Wild West” where the rules for managing non-human access are still being written. BeyondTrust, a company built on the premise that privilege is the true source of risk, is now positioning itself to secure these powerful new actors. The company’s Pathfinder Platform is designed to unify the security of human, machine, and now AI agent identities, applying principles of least privilege and continuous monitoring to entities that don’t have a human operator.

By using Claude Mythos Preview to harden its own platform, BeyondTrust is not just securing its products; it is building a more resilient foundation to manage the security of its customers' AI-driven ecosystems. This move acknowledges that in an era where AI agents can be both valuable assets and significant liabilities, governing their privileges is no longer an abstract problem but a core component of corporate strategy and cyber resilience.

A Coalition of Defense

The threats facing global infrastructure are too vast and complex for any single organization to handle alone. This is the central philosophy behind Project Glasswing, which has expanded from its initial 12 founding partners to a coalition of approximately 150 organizations across more than 15 countries. The initiative represents a new paradigm for cybersecurity: collaborative, AI-powered defense.

“The threats ahead are bigger than any one vendor, and the response has to be shared,” noted Marc Maiffret, Chief Technology Officer at BeyondTrust. “We are honored to stand with Anthropic and the other members of Project Glasswing, applying Mythos to our own code to further strengthen the security of the products our customers depend on, and doing our part in a defense no one can mount alone.”

This collaborative model, however, introduces its own set of challenges. The firehose of vulnerabilities being discovered by Mythos has shifted the industry’s bottleneck from discovery to remediation. Analysts note that the human capacity to triage, patch, and deploy fixes is now the primary limiting factor. While AI is exceptionally good at finding problems, the industry has yet to fully automate the solution, creating a potential backlog of known but unfixed vulnerabilities that could itself become a risk.

The Double-Edged Sword of Frontier AI

Anthropic’s decision to provide access to a model as powerful as Mythos, even within a controlled environment, is not without risk. The company has been transparent about the dual-use nature of this technology, privately warning officials that the proliferation of such capabilities could make large-scale cyberattacks more likely if they fall into the wrong hands. Project Glasswing is, in essence, an attempt to stay one step ahead in a race that Anthropic itself helped start.

By leveraging Mythos for defense, BeyondTrust and its fellow coalition members are engaging in a high-stakes effort to fortify the digital world's most critical systems before an adversary can use similar technology against them. The initiative underscores a fundamental truth of the modern era: identity is the new perimeter, and securing the privileges within that perimeter—whether held by a human administrator, a cloud service, or an autonomous AI—is the central challenge of our time. The success of this grand experiment in collaborative defense will have profound implications for the security and stability of the digital society we all depend on.