Anthropic recently unveiled Claude Mythos Preview, a general-purpose language model with heightened ability to identity cyber exploits. The model scored 93.9% on SWE-bench Verified and 83.1% on CyberGym. According to Anthropic, the model has already discovered “thousands of high-severity vulnerabilities, including some in every major operating system and web browser.”
While useful for cybersecurity purposes, cybercriminals would find the technology just as useful for leveraging vulnerabilities.
Educational Webinars, Videos & Podcasts: Receive cutting-edge insights and invaluable resources, empowering you to stay ahead in the dynamic world of security.
Empowering Content: At your computer or on-the-go, stay up-to-date when you receive our eNewsletters curated with the latest technology and services that address physical, logical, cyber and enterprise resilience.
Unlimited Article Access: Dive deep into the world of cybersecurity and risk management leadership with unlimited access to our library of online articles.