Anthropic: AI finds bugs faster than humans can fix them

anthropic's project glasswing reveals that claude mythos preview has uncovered 10,000 high-severity vulnerabilities in critical infrastructure software, with partners reporting a tenfold jump in discovery rates that exceeds their ability to verify and patch. companies including cloudflare and palo alto networks confirmed massive increases in patch volumes, while anthropic's scan of 1,000 open-source projects identified over 6,000 potential flaws with a 90% true positive rate. although the model successfully prevented a $1.5 million bank fraud, security risks are escalating because the standard two-week fix timeline is too slow for the influx of findings, leaving maintainers overwhelmed and forcing anthropic to withhold technical details to protect users.