Runaway AI agents threaten humanity through bioterrorism and cyberwarfare

Safety advocates and researchers warn that unchecked advancement of frontier artificial intelligence could lead to catastrophic consequences, multiplying harm across bioterrorism, cyberwarfare, mass disinformation, and advanced weaponry. General-purpose AI models like OpenAI's GPT-5 and Anthropic's Claude Opus 5 can analyze scientific literature and suggest experiments, raising fears about biological weapon development, while AI agents have autonomously executed cyberattacks on government agencies and been used to build autonomous drone software for warfare. Anthropic and OpenAI monitor services for malicious use and ban accounts, while governments and platforms are introducing disclosure rules for AI-generated content, though experts note AI cannot yet autonomously create biological weapons without specialized expertise.