Breaking

OpenAI launches dots as ChatGPT gains always-on agents for background work

October 2, 2026 at 07:49 EDT

AI Agents
Foundation Models
Sales & Business

OpenAI launched dots on September 29, bringing GPT-6 Astra-powered agents into ChatGPT that can keep working after a conversation ends. The rollout starts with one primary dot for eligible subscribers, while teams of dots and specialist organizational dots remain future plans or focused enterprise pilots.

September 29, 2026 · OpenAI

Dots move ChatGPT into always-on work

OpenAI's dots run in their own cloud computers, keep context between conversations, and return work for review. The first version starts with one primary dot while specialist and multi-dot setups remain bounded.

4,000+

apps through plugins

24/7

agent availability

primary dot at launch

91.8%

scope-change pass rate

Where scope risk showed up

In chained-task tests, flag rates rose when related intervening tasks doubled.

8.6%

five tasks

19.7%

ten tasks

More persistent task chains need tighter boundaries and review.

How a dot is meant to work

GoalUser defines the task

→

Cloud computerAgent works separately

→

Apps and toolsPermissions set access

→

ReviewSensitive actions pause

Best early fit

Bounded work that can be checked: calendar review, research drafts, content preparation, document review, customer-feedback triage and recurring admin.

Watch points

Persistent agents raise the stakes for permissions, sensitive actions, regional access, desktop-first setup and review of consequential work.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.