BlockBeats News, June 9th, Anthropic's latest test results show that its AI model Mythos has successfully exploited 18 of the 21 recently disclosed Windows kernel vulnerabilities, with a success rate of approximately 86%. Mythos is able to automatically generate executable network attack exploits based on publicly available vulnerability information, with an average cost of around $2,000 per successful exploit code.

Anthropic stated that Mythos has demonstrated the ability to transform public vulnerabilities into effective attack tools. The related research highlights the double-edged sword effect of advanced AI models in the field of cybersecurity, raising further concerns in the industry about the boundaries of AI capabilities and security governance.