Tenable integrates Anthropic Claude Mythos 5 into One exposure platform
Tenable Holdings (NASDAQ: TENB) has announced it will embed Anthropic's Claude Mythos 5 large language model directly into its Tenable One Exposure Management Platform, expanding an existing collaboration with the AI safety company under a programme called Project Glasswing. The integration marks a shift for that partnership, which previously focused on securing Tenable's internal code and infrastructure; the model will now be surfaced to customers as a cyber-reasoning layer within the commercial product.
The first feature to emerge from the expanded arrangement is Tenable One Adversary View. According to the company, it uses Claude Mythos 5 to analyse existing exposure data and reconstruct how a threat actor could traverse a network from an initial point of compromise through to critical systems. The capability is designed to show analysts the evidence behind each modelled step and recommend specific remediations that would break the attack chain. Tenable says Adversary View will be available to an initial set of customers in September, with further AI-powered capabilities planned for Q4 2026 and beyond.
The product case
The core problem Tenable is trying to solve is well understood by security operations teams: the volume of vulnerability and exposure data already collected is rarely the bottleneck. The harder challenge is correlating apparently unrelated findings into coherent attack paths and then ranking them by exploitability and potential blast radius. Attack-path analysis tools have existed for several years, but Tenable is positioning Claude Mythos 5's reasoning capabilities as a qualitative step forward in interpreting that graph-style data at cloud scale.
Eric Doerr, chief product officer at Tenable, said the integration would help customers "see their environments as an attacker would and identify the actions that can reduce risk most effectively," describing Adversary View as "just the beginning" of the planned innovation roadmap.
Tenable did not disclose pricing for Adversary View, whether it will be available across all Tenable One licensing tiers, or which Anthropic model version (API-hosted or on-premises) underpins the integration.
Market context
Tenable operates in a crowded exposure-management segment alongside CrowdStrike, Qualys, Rapid7 and Wiz, all of which have accelerated AI feature development over the past 18 months. The injection of frontier reasoning models into security workflows reflects a broader pattern: vendors are moving away from deterministic rule-based prioritisation engines and toward probabilistic, context-aware analysis that can handle the complexity of hybrid cloud and OT environments.
Anthropic's commercial partnerships have expanded steadily across enterprise verticals, with Claude models appearing in legal, financial services and now cybersecurity tooling. The Project Glasswing framing suggests a structured, multi-phase arrangement rather than a straightforward API licence, though Tenable has not disclosed commercial terms.
From a regulatory standpoint, security vendors deploying general-purpose AI models in agentic or semi-automated remediation workflows will face increasing scrutiny under the EU AI Act's high-risk classification criteria, particularly where outputs influence access-control or incident-response decisions. The UK's Cyber Security and Resilience Bill, currently progressing through Parliament, may impose additional transparency obligations on AI-assisted security tooling sold to operators of critical national infrastructure. Tenable serves more than 40,000 customers globally, including enterprises in regulated sectors where those requirements will apply earliest.
The September availability window for Adversary View is narrow; Tenable's forward-looking statements caution that delays in AI product rollouts and customer adoption challenges are live risks.