Project Glasswing has uncovered 10,000 vulnerabilities: Anthropic
Summary
Anthropic's Project Glasswing initiative, utilizing its Claude Mythos Preview AI model, has discovered an estimated 10,000 critical or high-severity vulnerabilities in software offerings. The AI model demonstrated a capability to find vulnerabilities surpassing most human experts, leading to the scanning of over 1,000 open-source projects.
IFF Assessment
The article highlights an AI-powered initiative successfully identifying a large number of critical vulnerabilities, which aids defenders in patching and securing systems.
Defender Context
This initiative demonstrates the growing capability of AI in identifying software vulnerabilities, a trend defenders need to monitor. It also underscores the ongoing challenge for open-source maintainers in managing a deluge of bug reports, potentially impacting the speed of patching critical flaws.