Anthropic has revealed that its Project Glasswing cybersecurity initiative identified more than 10,000 high- and critical-severity vulnerabilities in widely used software systems within one month of launch. The project gives around 50 trusted partners early access to Claude Mythos Preview, an advanced AI model designed to autonomously detect software vulnerabilities before cybercriminals can exploit them.
According to Anthropic, over 6,200 vulnerabilities were classified as high or critical severity across more than 1,000 open-source projects. Security researchers later confirmed 1,726 valid vulnerabilities, including 1,094 categorized as critical or high-risk flaws. One major finding involved a severe WolfSSL vulnerability that could allow attackers to forge certificates and impersonate legitimate services.
The initiative has already contributed to 97 security patches and 88 advisories. Anthropic noted that AI-assisted vulnerability discovery is significantly increasing the number of security fixes released by technology companies globally.
Cybersecurity platform XBOW described Mythos Preview as a major advancement in AI-powered threat detection, capable of identifying complex vulnerabilities and attack chains with high accuracy. Beyond software security, the AI system reportedly helped a banking partner prevent a fraudulent $1.5 million wire transfer linked to account compromise and spoofed communications.
Anthropic urged organizations to strengthen cybersecurity defenses by accelerating patch deployment, enabling multi-factor authentication, improving logging systems, and hardening network configurations. The company also launched a Cyber Verification Program that allows approved researchers to use its AI models for penetration testing, vulnerability research, and red-team operations while maintaining safeguards against misuse.
See What’s Next in Tech With the Fast Forward Newsletter
Tweets From @varindiamag
Nothing to see here - yet
When they Tweet, their Tweets will show up here.
