What Anthropic Announced
Anthropic said on August 21, 2026 that Claude Mythos 5, its most capable frontier model, now powers Claude Security for every Claude Enterprise customer.
The company had run Claude Security in public beta since April 30, 2026, giving vetted defenders early access to frontier-model scanning before folding the capability into the standard Enterprise tier roughly four months later.
How The Billing Change Works
Anthropic said scans run through Claude Security are now billed as standard token usage under a customer's existing plan, with no separate add-on and no separate model access required.
Every finding carries a CWE classification, a confidence rating, and a severity rating, plus a suggested patch that Anthropic said still needs a human reviewer before it ships, and Enterprise admins can turn the feature on from their own admin console.
| Claude Security public beta launched | April 30, 2026 |
|---|---|
| Mythos 5 rollout to Enterprise announced | August 21, 2026 |
| Billing model | Standard token usage, no separate add-on |
| Defender Advantage Fund | 35 million dollars in Claude credits |
The NIS2/DORA Cost-Calculus
This billing change resets the cost-benefit math behind NIS2 and DORA vulnerability-management duties for any mid-size EU or UK firm that was postponing automated scanning because of headcount cost.
Until now, running a frontier model against a live codebase for continuous vulnerability scanning meant either hiring a security engineer to operate it or paying for a dedicated security product on top of whatever AI plan the company already held, and Anthropic has now folded that capability into the Enterprise subscription itself.
For an owner who reports NIS2 or DORA compliance status to a board or a regulator, that turns automated scanning from a discretionary hire into a line item already sitting inside an existing AI contract, which changes who gets asked to explain why scanning still is not running.
The Open-Source Supply-Chain Angle
The 35 million dollar Defender Advantage Fund is a second signal worth reading past the headline figure: Anthropic is subsidizing the same open-source security tooling that its own Claude Security scanner depends on.
Anthropic said the fund, structured as Claude credits, targets organizations patching live vulnerabilities in widely used open-source projects, automating scanning and patching, and pursuing more ambitious security work, which matters to any firm relying on the same open-source components without funding their upkeep.
What Enterprise Customers Should Do Now
Enterprise customers do not need to buy a separate product to get frontier-model scanning, because Anthropic said the feature switches on from the existing admin console.
Firms outside the Enterprise tier, or security teams wanting deeper access, can apply through Anthropic's Cyber Verification Program, and Anthropic pointed to its Project Glasswing partnership with the U.S. government as the model for how it extends early access to critical-infrastructure defenders.
Read next: OpenAI Confirms Astra Pause, Urges AI Pacing | Anthropic Posts $11.5B Quarter, Its First Profit
