At its annual DevDay conference in San Francisco, OpenAI announced the launch of “Dots,” a new category of proactive, always-on artificial intelligence agents engineered to perform complex, multi-step tasks continuously with minimal human supervision. Designed to operate in the background across cloud virtual machines, web browsers, and local desktop environments, Dots mark a major push by the ChatGPT creator toward fully autonomous digital assistants capable of anticipating user needs and executing multi-application workflows.
Read: Bandwidth Blog & Smile 90.4FM Tech Tuesday: Rogue AI!
Unlike traditional conversational models that respond strictly to explicit prompts, Dots are engineered to independently monitor workflows, identify necessary actions, and carry out background operations. The agents leverage OpenAI’s broader plugin ecosystem to interact with over 4,000 third-party software applications. Through new enterprise integrations with platforms such as Slack and Microsoft Teams, users can communicate with their assigned Dot outside of standard OpenAI environments like ChatGPT, Codex, and Work applications.
Dots can operate within dedicated cloud environments or connect directly to a user’s local workstation to control applications on their behalf. In practical applications, an agent can automatically detect unbilled project milestones to generate draft invoices, continuously synthesize incoming laboratory research data into academic manuscripts, or aggregate user feedback for software applications to draft code bug fixes autonomously. Users retain real-time visibility into an agent’s activities, with the ability to inspect its virtual computing workspace or interrupt tasks at any moment.
Interactions with Dots occur via text messaging or voice calls across desktop, web, and mobile interfaces, with context maintained seamlessly across all platforms. In addition to receiving direct commands, Dots can proactively message users with status updates, clarification questions, or requests for authorization before completing sensitive actions.
To address ongoing safety and security concerns surrounding autonomous agent operations, OpenAI has incorporated strict permission boundaries into the platform. When logging into web services, Dots utilize encrypted saved credentials without exposing raw passwords to the underlying language model. Furthermore, background research and monitoring capabilities are bound by default read-only restrictions, preventing agents from altering application content, dispatching unauthorized communications, or modifying local system settings without explicit user consent. Administrative controls within ChatGPT allow users to establish custom operational rules specifying which actions require manual approval.
OpenAI is initiating a phased commercial rollout of Dots starting today for Pro, Business, Premium, and Enterprise subscribers in select international markets. Account holders in eligible tiers receive access to one primary, customizable Dot included with their subscription, featuring unlimited conversational queries within ChatGPT, alongside standard usage limits across Codex and Work applications. OpenAI plans to expand the platform in future updates to allow enterprise clients to deploy and coordinate multi-agent teams.
