On 29 September 2026, OpenAI published a DevDay product post introducing dots: always-on agents powered by GPT‑6 Astra. Each dot has its own cloud computer and browser, can connect to more than 4,000 apps through OpenAI’s plugin ecosystem, and can reach users in ChatGPT, Slack, and Microsoft Teams. A same-day safety post explains how permissions, background research, and action review keep consequential decisions with the user. This brief focuses on those checkable product boundaries rather than listing every DevDay announcement.
Always on—not another chat thread
OpenAI describes dots as “a whole new way to work with AI”: agents that learn what matters to you, keep working on your behalf, and give time and attention back. Unlike one-shot Q&A, a dot can run with a project while juggling others. Users can open the dot’s computer at any time to inspect its work, and may also grant permission to connect a personal laptop so the agent can help beside local files or installed tools. Internal examples include dots investigating a bug as soon as it appears in Slack, turning a new design into a working app while the team stays with customers, and keeping a planning cycle aligned to a deadline. An early tester’s dot noticed a missed invoice, prepared it, and sent it after approval.
Where they run, and who can start
Users can message or call dots in ChatGPT on desktop, web, and mobile, and can also message them in Slack and Teams, with texting coming soon. OpenAI says dots are beginning to roll out to Pro and Business Premium in eligible markets. Enterprise, Edu, and Healthcare users can try the beta when a workspace admin enables it; it is off by default. The first dot is included in Pro or Business Premium at no extra cost, with an allowance for deeper work and extended limits in the first month after launch. Conversations with a dot do not count toward ChatGPT usage limits; tasks the dot starts or manages in Codex or ChatGPT Work still count under those products’ usual limits.
Background help: read-only proactive research
When the user is not actively engaged, a dot can look for ways to help in the background—proactive research. Those tasks run in each dot’s cloud environment with read-only tools on permitted connected sources: they cannot directly send messages to other people, change content in connected apps, or control a browser or desktop. Findings are saved as private notes for that dot; any follow-up action still follows the usual rules and checks. OpenAI states it does not train models directly on those background research threads or their notes; if a note later enters an eligible conversation or task, it may be used for model improvement depending on the user’s settings.
Permissions, Custom Rules, and Auto-review
Users choose which apps dots can access through existing ChatGPT app and plugin controls. Custom Rules can allow actions, require approval, or block them, but cannot remove mandatory safety requirements. Before sending a message or sharing a file, dots are taught to seek authorization that covers the information and the type of recipient; more sensitive data needs more specific recipients. Purchases with cards already saved on a merchant site require user approval. Steps such as permanently deleting data, installing or running software from an unrecognized source, or granting new security-sensitive access need confirmation each time. Changing a password or transferring money between financial accounts must be handed back to the user, even if the dot helps with surrounding work.
Before actions such as sending email or changing files, a separate system called Auto-review checks planned steps against the user’s instructions, Custom Rules, and safety requirements. If it blocks a step, it tells the dot why; the dot may ask the user, try a permitted alternative, hand a sensitive step back, or stop. OpenAI says the controls that enforce Auto-review sit outside environments dots can change. Activity View in the desktop app shows ongoing and delegated work so users can add context, redirect, or stop a task.
Credentials and workspace isolation
Each dot works on its own cloud computer; the user’s machine and contents stay separate unless the user chooses to connect them. For supported website sign-ins, secure sign-in and saved-password flows keep passwords out of the model’s context: the model pauses while credentials go through an encrypted path into the browser environment, then work resumes. Sandboxing limits what code and tools a dot can reach. Systems that coordinate work and enforce key safeguards are separated from the environments where dots run code, so a mistake cannot turn off required checks. Monitoring can pause work and surface a warning if safety concerns appear.
For ChatGPT Business, Enterprise, and Edu workspaces, content is not used to improve models by default; on personal plans, users control whether dots’ conversations and work are used via settings such as “Improve the model for everyone.” OpenAI states that dots can still make mistakes, so consequential work should be reviewed; more detail is in the dots safety post and system card.
Specialist dots for organizations—and reading limits
Beyond a personal primary dot, OpenAI previews specialist dots: company-provisioned agents with their own identity, credentials, and access to systems of record for well-defined roles such as procurement, invoice processing, email marketing, customer support, and commercial contracting. Rollout starts with focused enterprise pilots. OpenAI is also working with Microsoft to integrate specialist dots with enterprise governance and security controls in Agent 365. Over time the company envisions teams of dots working together; today’s materials still center on a first primary dot and pilots.
The same-day DevDay Recap also covers GPT‑6.1 Sol, Ultrafast, Codex cloud and security tooling, ChatGPT Space, plugin extensions, and more. This brief keeps the focus on dots’ always-on form factor and checkable controls. Eligible markets and admin switches determine who can enable the product; readers should verify against their OpenAI account and official docs.
Primary sources
- OpenAI: Introducing dots (2026-09-29)
- OpenAI: How we build safety, security, and privacy into dots (2026-09-29)
- OpenAI: DevDay 2026 Recap (2026-09-29; event context)
