The release marks a significant expansion of Anthropic's "Project Glasswing" initiative, which previously limited access to this high-tier model to a small group of organizations protecting critical infrastructure.
The deployment aims to give defenders a technological edge over malicious actors by providing high-level reasoning for identifying software weaknesses.
To mitigate the risk of the model being used for offensive "dual-use" purposes—meaning capabilities that could help both defenders and attackers—Anthropic is restricting access to specific defensive outputs rather than providing an open-ended chat interface.
By integrating the model into specialized security tools, the company intends to help teams triage alerts and remediate threats faster than traditional manual methods.
Beyond the Enterprise tool, Anthropic is launching the $35 million Defender Advantage Fund (0xDAF) to provide model credits to organizations securing open-source software.
The company also plans to expand its Cyber Verification Program, which grants vetted security professionals reduced safeguards on models like Opus and Sonnet, with Mythos-class access expected to follow.
These programs are designed to automate the scanning of widely used public projects that often lack the resources to defend against sophisticated modern attacks.