“AI models like Mythos can be used by offensive actors to find external flaws, allowing them to daisy-chain vulnerabilities and penetrate infrastructure.”