Autonomous AI cyberattack campaign using DeepSeek and the Hermes Agent framework attacked 460-plus targets after a Chinese hacker found Claude and OpenAI’s safety controls blocked offensive use — Unit ...