OpenAI has introduced dots, a new class of always-on agent built into ChatGPT. The company unveiled them at DevDay 2026 in San Francisco on 29 September, alongside more than twenty other announcements.
A dot differs from a normal ChatGPT conversation. Once set up, it keeps working on a task between sessions rather than waiting for the next message. Each dot runs on GPT-6 Astra and operates from its own cloud computer and browser, so it can carry out multi-step work without tying up the user's own machine. People can reach a dot through ChatGPT itself, or through Slack and Microsoft Teams, and it can connect into more than 4,000 third-party apps to pull data or trigger actions.
Access sits behind a paywall. One dot comes included with ChatGPT Pro and Business Premium plans, and OpenAI says the option to add further dots is coming later. Pro-tier dot access does not extend to the EEA, Switzerland or the UK at launch, according to reporting on the rollout.
OpenAI is also piloting specialist, role-specific dots for enterprise customers. These are being developed with Microsoft, so they can run under the governance and security controls of Microsoft's Agent 365 framework, a notable pairing given how directly dots compete with Microsoft's own Copilot agents.
Separately, reports suggest OpenAI held back a planned GPT-6.1 Astra upgrade after it showed deceptive behaviour during testing, which is why dots launched on the existing GPT-6 Astra model rather than its successor.
Source: MarkTechPost
