Anthropic Launches Claude Security on Mythos 5 and $35M Open-Source Fund
Anthropic has released Claude Security into public beta for Enterprise customers, utilizing its Claude Mythos 5 model for GitHub vulnerability scanning. The company also introduced the $35 million Defender Advantage Fund to provide Claude credits for securing open-source software.
Impact: High
Why it matters
Enterprise engineering teams can run automated cross-file vulnerability audits on GitHub codebases, while open-source maintainers can apply for dedicated AI security grants.
TL;DR
- 01Claude Security scans powered by Claude Mythos 5 are available in public beta for Claude Enterprise customers.
- 02The scanning interface restricts model output to structured vulnerability reports and patches, avoiding open prompt execution.
- 03Anthropic launched the $35M Defender Advantage Fund to provide Claude credits for open-source security maintainers.
Key facts
- Scanner Model
- Claude Mythos 5
- Open Source Grant
- $35 million (0xDAF)
- Access Tier
- Claude Enterprise public beta
- Vulnerability Scope
- Memory corruption, injection, auth bypass, cross-file logic
Claude Mythos 5 in Claude Security
Anthropic has deployed Claude Mythos 5 within Claude Security, now available in public beta for Claude Enterprise organizations. The scanner links directly to GitHub repositories, analyzing data flows across multiple files and reading Git history to detect severe vulnerabilities such as memory corruption, injection flaws, and authentication bypasses.
Findings include:
- Common Weakness Enumeration (
CWE) classification - Severity and confidence ratings
- Recommended code remediation patches
- Direct webhook exports to Slack, Jira, CSV, or Markdown
Constrained Interface Architecture
To prevent the underlying cybersecurity model from generating offensive exploits, Claude Security does not provide an open prompt interface. Mythos 5 operates in the background, returning static audit findings. Interactive code editing and patch application take place in Claude Code using the standard models already configured on the organization's plan.
$35M Defender Advantage Fund (0xDAF)
Anthropic also announced the Defender Advantage Fund (0xDAF), allocating $35 million in Claude credits to developers and organizations responsible for securing critical open-source software and infrastructure.
✓ When to use
- Deep architectural and data-flow security audits across GitHub codebases.
- Securing open-source repositories eligible for the 0xDAF credit grants.
✕ When NOT to use
- Interactive conversational development, as Mythos 5 is not exposed via direct chat or prompt endpoints.
- Organizations on Claude Pro, Max, or Team tiers without Enterprise agreements.
What to do today
- Enterprise administrators can enable repository audits in the admin console via claude.ai/security.
- Open-source maintainers can apply for Claude API security credits through the Defender Advantage Fund.
Sources