460/69 Tuesday, August 25, 2026
Anthropic has officially announced expanded access to the cybersecurity capabilities of its advanced AI model, Mythos 5, for defenders through integrations with cybersecurity partners, updates to Claude Security, and the expansion of its Cyber Verification Program. The company also introduced the Defender Advantage Fund (0xDAF), a USD 35 million fund in the form of Claude usage credits, to support open-source software developers in fixing vulnerabilities and improving the security of critical infrastructure. This initiative builds on Project Glasswing, aiming to provide defenders with tools to detect and remediate vulnerabilities before such capabilities can be misused.
In terms of deployment, Anthropic has designed access to Mythos 5 with an emphasis on limiting direct interaction to reduce the risk of misuse. Users will access the model through interfaces provided by partner security tools, such as security operations center (SOC) processing systems, incident response platforms, and threat detection tools used by hospitals, financial institutions, utilities, and software supply chains. In addition, for Claude Enterprise customers, Claude Security has been upgraded to use Mythos 5 to scan source code, identify vulnerability categories based on the CWE standard, assess confidence and severity levels, and recommend patches. However, actual code changes must still be reviewed and approved by qualified human experts before implementation.
For cybersecurity organizations and administrators, Anthropic is allowing security teams to apply for the Cyber Verification Program to receive adjusted safety restrictions for testing, triaging, and validating vulnerabilities using Claude Opus and Sonnet models, before expanding to Mythos-level model access in the next phase. Open-source projects can also apply for support from the 0xDAF fund to fix vulnerabilities and build repeatable security review processes. Developers and IT administrators should review their source code auditing processes, improve software supply chain management, and consider adopting defensive AI tools as part of a layered monitoring and security strategy.