Current outcome
Anthropic launched OSS Scanner, an opt-in service that generates vulnerability reports for open-source projects using its strongest frontier models, prompting applications from crypto projects including Nethermind and ZEUS.
Progress timeline
3 material updates- #01
Anthropic给Claude开「红队权限」：安全机构可做授权渗透测试
Anthropic expanded its Cyber Verification Program into three tiers — Defense, Red Team, and Specialized — granting vetted security organizations reduced cybersecurity guardrails on Claude Opus 5.5, Sonnet 5.5, and Mythos 5.1, with the Red Team tier allowing authorized penetration testing and reportedly completing 34 of 50 complex attack tasks that were fully blocked on the standard model.
Source evidence: theblockbeats
- #02
Launching an opt-in vulnerability-finding service for open-source software - Anthropic
Anthropic announced OSS Scanner, an opt-in service that uses its models to run thorough, periodic security scans of open-source software repositories and alert projects to potential vulnerabilities at no cost. According to the source, open-source projects that opt in could be alerted about possible security issues sooner; the trade-off is described as stemming from how OSS Scanner works. The service is described as free to participating open-source projects and uses Anthropic's strongest models for periodic scanning; participation is opt-in rather than automatic, and the company maintains a public repository for the service.
Source evidence: Anthropic News
- #03
Anthropic免费用Claude扫开源漏洞，已发现超2.9万处疑似漏洞
Anthropic launched the Cyber Mission program with a free Claude-powered open-source vulnerability scanner and a critical-infrastructure defense initiative, reporting more than 29,000 suspected vulnerabilities uncovered in open-source software over the past six months in partnership with 11 companies.
Source evidence: theblockbeats · Cointelegraph