How AI is revolutionizing—and endangering—global cybersecurity
Anthropic's AI model, Mythos, can discover and exploit software vulnerabilities at a level comparable to skilled humans, raising concerns about its potential misuse. The company has shared Mythos with over 40 tech firms to help them defend against cyber threats.
In 2016, the Pentagon's research agency hosted a hacking competition won by a computer system called Mayhem. Since then, AI has advanced significantly, with Anthropic's Mythos model capable of discovering and attacking software vulnerabilities like skilled humans. Anthropic has shared Mythos with over 40 tech companies to help them defend against attackers. However, some analysts argue that existing open-source AI models can already find and exploit vulnerabilities. Companies like Aisle have used AI to develop methods for autonomously detecting potential vulnerabilities, identifying security weaknesses in vital digital infrastructure.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.