A new Anthropic model found security problems ‘in every major operating system and web browser’
Why it matters: Project Glasswing aims to autonomously identify and patch thousands of high-severity vulnerabilities in major operating systems and web browsers.
- Anthropic is launching Project Glasswing, a cybersecurity partnership with major tech companies, to autonomously flag system vulnerabilities.
- Claude Mythos Preview, a new general-purpose AI model not publicly released due to security concerns, has found "thousands of high-severity vulnerabilities" in every major operating system and web browser.
- Newton Cheng, Anthropic's cyber lead, stated the model provides cyber defenders a "head start" against adversaries, with access restricted to "defensive security" partners.
- The model identified vulnerabilities and developed exploits "entirely autonomously, without any human steering," according to Anthropic's blog post.
- Dianne Penn, head of product management at Anthropic, attributed a recent data leak of Mythos Preview's existence to human error, not software vulnerabilities.
- Anthropic will commit up to $100 million in usage credits and $4 million in direct donations to the Linux Foundation and Apache Software Foundation to subsidize the model's use for its partners.
Anthropic has unveiled Project Glasswing, a cybersecurity initiative powered by its new Claude Mythos Preview AI model, which has autonomously identified thousands of high-severity vulnerabilities across major operating systems and web browsers. This model, developed in partnership with tech giants like Nvidia, Google, and Apple, is currently restricted to "defensive security" partners to prevent misuse by adversaries. The company is subsidizing its use with up to $100 million in credits and donations to open-source foundations.
