The security effort expands upon a 3 tier clearance system for Claude Mythos 5.1, Opus 5.5, and Sonnet 5.5, which is available to vetted security professionals for authorized penetration testing and malware analysis. Anthropic accelerated these protections after Project Glasswing partners identified 129,000 verified vulnerabilities, including 33,000 rated as high or critical severity. The startup also noted that rival Zhipu AI's open GLM-5.3 model could generate exploits in 50 of 410 attempts, nearly matching the capabilities of Anthropic's Mythos Preview.
Tuesday, October 20, 2026
Story
Anthropic Launches First AI Bug Scanner to Protect Open Source Software
Cybersecurity12 days ago
The AI startup is introducing a free, opt-in tool to help open-source software maintainers identify security flaws and distribute fixes more rapidly via its new Anthropic Cyber Mission. This initiative includes the release of the OSS Scanner, which uses the Claude Mythos model to detect vulnerabilities and provide proofs of concept. As part of the broader mission, the company is also deploying a Critical Infrastructure Defense Program with 11 founding partners, including CrowdStrike, Palo Alto Networks, and Rockwell Automation, to protect utility and transportation grids.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.