OpenAI has introduced dots — always-on AI assistants built on GPT-6 Astra designed to run continuously on a user’s behalf. Each dot runs on its own dedicated cloud computer, includes an integrated browser, and can connect to an ecosystem of plugins that link it to over 4,000 third-party apps. OpenAI positions dots as persistent collaborators that learn from feedback, manage projects autonomously, and surface progress in ways tailored to individual workflows.
Dots are available through ChatGPT on desktop, web and mobile, and are also accessible in Slack and Microsoft Teams. Interaction options include messaging and calling a dot, receiving progress updates, asking project-related questions, and providing feedback. A key design goal is cross-channel continuity: a conversation begun in ChatGPT can continue in Slack or Teams without losing the underlying project context, allowing tasks to move between environments while the dot retains relevant information.
Each dot can open and manage multiple projects simultaneously. Users can inspect a dot’s dedicated cloud computer to review its work at any time and may grant permission for a dot to access their laptop so it can assist directly alongside them. In early internal testing, OpenAI reports dots performing a range of tasks such as investigating bugs in Slack, transforming a design into a working app, and coordinating planning cycles — examples that illustrate how a dot might act proactively to keep work moving.
OpenAI has built a suite of safety, privacy and control features around dots. Because each dot runs on a separate cloud computer, a user’s personal device and files remain isolated unless the user explicitly connects them. For supported website sign-ins, dots may use saved credentials without exposing passwords to the model. When idle, dots can perform background activity described as “proactive research,” but during such work connected apps are limited to read-only access: a dot cannot send messages, modify app content, or control a browser or computer without explicit permission.
Monitoring systems are in place to detect potentially harmful behavior, and OpenAI can pause or stop a dot’s activity if safety concerns arise. Users control which apps a dot can access and manage permissions through ChatGPT’s controls. Dots ship with built-in rules that determine when they may act independently and when they must request approval. Administrators and users may also set Custom Rules to allow, require approval for, or block specific actions. An Activity View provides visibility into background work so users can follow progress and intervene as needed.
To limit risky or consequential actions, dots employ an auto-review mechanism that evaluates actions likely to affect accounts or share information. That system checks proposed actions against the user’s instructions, Custom Rules and safety requirements, and it categorizes work into items that can proceed automatically, those requiring approval, and tasks the user must perform directly. OpenAI notes that certain sensitive operations — such as changing a password — always remain with the user, and cautions that dots can still make mistakes, so consequential work should be reviewed.
OpenAI is also offering specialist dots for organizations that need agents with defined responsibilities and integrations. Specialist dots are provisioned with distinct identities, credentials and access to company systems of record. Early internal testing inside OpenAI covered areas including procurement, invoice processing, email marketing, customer support and commercial contracting. The company has started focused enterprise pilots and plans direct engineering collaboration with organizations to specify each specialist dot’s responsibilities, tooling and review workflows. OpenAI is working with Microsoft to integrate specialist dots with Microsoft Agent 365 governance and security controls so businesses can manage dots via existing Microsoft tools.
Dots began rolling out to eligible markets for Pro and Business Premium plans, with Enterprise workspaces able to try a beta when enabled by a workspace administrator. Each Pro and Business Premium plan includes a primary dot at no additional cost. OpenAI has set an initial allowance for deeper work and extended limits for the first month after launch. Conversations with a dot do not count toward ChatGPT usage limits, though tasks that invoke Codex or ChatGPT Work are counted against standard usage limits. To get started, users create a dot in the ChatGPT desktop app or a desktop browser, connect apps and let the dot introduce itself; OpenAI says it will continue expanding availability and refining dots based on user feedback.
OpenAI’s dots introduce a persistent, account-linked assistant model intended to move beyond one-off prompts toward continuous collaboration. The company emphasizes layered controls, separations between cloud-run dots and personal devices, and administrative tooling for enterprises — underscoring that while dots can take initiative, oversight and safety remain core to their design.
Source: Read the original source
Leave a Reply