05/28/2026
🤖 Anthropic says its unreleased AI model, Claude Mythos, has already found over 10,000 critical software vulnerabilities just one month after launch as part of its cybersecurity initiative, Project Glasswing.
🔎 The AI has reportedly helped companies dramatically increase the rate at which they discover security flaws. Cloudflare alone found around 2,000 bugs, including 400 classified as high or critical severity, while Mozilla said it found vulnerabilities in Firefox at around 10x the previous rate.
💻 Anthropic also used the model to scan 1,000 open-source projects, where it reportedly identified more than 6,200 high and critical vulnerabilities.
⚠️ Despite the success, Anthropic says it has not released Mythos to the public because current safeguards are not strong enough to prevent misuse. The company warned that a model capable of finding vulnerabilities this effectively could also potentially be abused for cyberattacks.
🏢 The system is already being tested with major organisations including Apple, Google, Microsoft, AWS, JPMorgan, NVIDIA and Palo Alto Networks.
🌍 It’s a glimpse into the future of cybersecurity — where AI may become powerful enough to find software flaws faster than humans ever could, potentially reshaping both digital defence and cyber warfare.
Follow for interesting stories, amazing places, and incredible people 💫