OpenAI’s Dots Are Always-On AI Agents—and Its Answer to Meta’s Muse

Your AI assistant now proactively acts on your behalf, moving beyond simple chat to manage tasks without you needing to initiate every step.

OpenAI has launched "Dots," an "always-on" AI agent designed to proactively complete tasks for users. These agents, powered by OpenAI's GPT-6 Astra model, continuously monitor the web and can integrate with connected applications to learn user preferences. In demonstrations, Dots assisted with dinner planning by suggesting and offering to order food, and helped launch a new website. Users can interact with Dots via ChatGPT, Slack, and Microsoft Teams, with an option for Pro users to join a waitlist for iMessage and RCS integration. The Dots service begins today for ChatGPT Pro subscribers, costing $100 per month, with plans for eventual wider rollout and support for multiple agents. OpenAI is also targeting enterprise clients with "specialist Dots" for specific professional tasks, signaling a move to embed AI agents further into workplace workflows. Safety features include explicit approval requirements for sensitive actions and a Custom Rules tool for user-defined boundaries. Users are advised to exercise caution regarding potential security and privacy risks, especially if their account settings permit model training on their data.

The introduction of always-on AI agents like OpenAI's Dots signifies a pivotal shift from reactive chatbots to proactive digital assistants. This move directly challenges competitors like Meta's Muse by offering a more deeply integrated and autonomous user experience that continuously works in the background. The agents' ability to pull context from connected apps and learn preferences suggests a future where AI is not just a tool, but a constant, personalized collaborator.

This development signals a significant push towards AI agents becoming integral components of both personal and professional workflows, particularly for enterprise clients seeking specialized AI support. The emphasis on user control through explicit approval for sensitive actions and custom rules highlights the industry's ongoing struggle to balance AI capability with user trust and security. As more personal data is integrated, the inherent privacy and security risks associated with these advanced agents will become a critical area for users and developers to navigate.