OpenAI's Dots Demand Human Approval Even for Agent-to-Agent Communication

OpenAI has launched Dots, always-on AI agents that pursue user-defined goals continuously in the background with strict human oversight—a philosophy so fundamental that product lead Alexander Embiricos will not even have a meeting with another AI agent without explicit approval. Announced September 29, 2026, at DevDay and powered by GPT-6 Astra, Dots operate on dedicated cloud infrastructure and connect to over 4,000 applications, marking a significant shift from reactive chatbots toward proactive autonomous agents in everyday workflows.