A Major Shift in AI-Powered Cybersecurity: Anthropic Expands Access to Mythos 5 and Unveils $35M Open Source Fund
In a significant move that could revolutionize the way organizations approach cybersecurity, AI giant Anthropic has announced plans to expand access to its advanced AI models, specifically Claude Mythos 5, while keeping guardrails in place to prevent misuse. This decision marks a major shift in the company’s strategy, as it seeks to provide more defenders with access to the powerful capabilities of its AI models without compromising security.
The expansion builds on Project Glasswing, launched earlier this year, which gave a select group of organizations early access to Claude Mythos Preview and its successor, Mythos 5. These models are designed to help defenders find and fix vulnerabilities before they can be exploited by malicious actors. Anthropic’s goal is to provide a safety net for organizations that need to use AI-powered cybersecurity tools without exposing themselves to the risks associated with direct interaction.
To achieve this balance, Anthropic is working with cybersecurity partners to integrate Mythos 5 into existing security operations, incident response, and detection tools used by teams protecting critical infrastructure. The integration will allow end-users to work through purpose-built interfaces that run the model in the background, returning only specific defensive outputs, such as a list of suggested patches or a security alert.
One notable development is the expansion of Claude Security, which now runs its codebase scans on Mythos 5. Scans surface each finding with detailed information, including CWE categories, confidence and severity ratings, and suggested fixes. However, any fix still requires human approval before deployment to prevent potential misuse.
In a separate move, Anthropic has launched the Defender Advantage Fund (0xDAF), committing $35 million in Claude credits toward organizations that help open source maintainers secure their projects. This fund builds on previous support provided through Project Glasswing, including direct donations and coordinated efforts with other security initiatives like Akrites and Gold Eagle.
The Cyber Verification Program, which already provides reduced safeguards for authorized security work, will also expand to cover broader dual-use capabilities on Claude models. Anthropic is encouraging security teams to apply now for reduced safeguards on Opus and Sonnet, with further details expected in the coming weeks.
This significant development marks a major step forward in AI-powered cybersecurity, as Anthropic seeks to balance access to powerful tools with the need for security and safety. As organizations continue to face evolving threats, this expansion of access to Mythos 5 and the launch of the Defender Advantage Fund offer new opportunities for defenders to stay ahead of malicious actors.
To take advantage of these developments, security teams should consider applying to the Cyber Verification Program now for reduced safeguards on Opus and Sonnet. Additionally, organizations can explore partnerships with Anthropic or its partners to integrate Mythos 5 into their existing cybersecurity tools. By doing so, they can harness the power of AI-powered cybersecurity without compromising their own security posture.
Source: SecurityWeek — 2026-08-24