The cybersecurity firm deepens its Anthropic partnership with a new adversary simulation feature powered by frontier AI reasoning
Tenable Holdings is embedding Anthropic’s Claude Mythos 5 directly into its flagship cybersecurity product, betting that frontier AI models can fundamentally change how organizations spot and neutralize threats before attackers exploit them.
The integration, announced on September 8, 2026, will bring AI-driven exposure analysis and attack-path discovery into the Tenable One Exposure Management Platform. The first concrete feature shipping under this effort is called Tenable One Adversary View, which uses Claude Mythos 5 to reconstruct how an attacker could move from an initial foothold all the way to an organization’s most critical systems.
What Adversary View actually does #
The feature analyzes exposure data across an organization’s entire attack surface and identifies the specific chains of vulnerabilities an adversary could exploit in sequence. Rather than handing security teams a flat list of thousands of individual issues ranked by severity score, it maps realistic attack paths from entry point to crown jewels.
Eric Doerr, Tenable’s Chief Product Officer, framed the capability as bringing “frontier cyber reasoning” into Tenable One.
The Project Glasswing backstory #
This isn’t Tenable’s first dance with Anthropic. The integration expands a collaboration that began under something called Project Glasswing, an initiative launched in mid-2026 focused on deploying advanced AI models specifically for cyber defense applications.
The partnership started with a more narrow scope. In May 2026, Tenable integrated Anthropic’s Claude Compliance API for AI governance purposes, essentially using AI to help manage AI-related risk. Around the same time, the company deployed its own proprietary Hexa AI model.
Initial customer availability for Adversary View is targeted for September 2026, with additional upgrades planned for the quarters that follow.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our