OpenAI gives you an always-on AI coworker

For the past few years, using AI meant babysitting a text box. You put in a prompt, waited for an answer, and the moment you closed your browser tab, everything ground to a halt.

OpenAI introduced dots, an always-on autonomous agent powered by GPT-6 Astra. The big shift here is autonomy: instead of just another chatbot waiting for your next instruction, a dot acts more like a persistent digital coworker that runs 24/7 in the background.

A persistent coworker with its own cloud PC

Unlike standard ChatGPT sessions, each dot comes provisioned with its own cloud computer and browser. Powered by an ecosystem of over 4,000 plugins, it can interact with your everyday workplace software to investigate bugs, write code, run tests, and prepare pull requests while you sleep.

You can reach your dot via ChatGPT on desktop or mobile, or ping it directly inside Slack and Microsoft Teams. Voice calls are supported right away, and text messaging is slated to arrive down the road. It can juggle multiple threads simultaneously without asking you to open new chats. In one early test case, a dot noticed its user had forgotten to invoice a publication, drafted the document automatically, and queued it up for a quick one-click sign-off.

OpenAI is currently rolling out dots to Pro, Business Premium, and Enterprise subscribers in eligible markets. The first dot is bundled into existing subscriptions at no extra charge, complete with expanded task limits during the first month. Early coverage points out that "eligible markets" largely leaves out the European Union for now due to local regulatory hurdles.

Putting guardrails on background agents

Letting an autonomous model wander around your apps around the clock naturally brings serious security concerns. OpenAI put several concrete boundaries in place to prevent rogue behavior.

For starters, each dot runs on an isolated cloud machine separated from your local laptop unless you explicitly grant device-sharing permissions. When logging into third-party accounts, dots can leverage saved passwords without exposing the credentials to the model's weights.

When left idle, dots enter a mode called "proactive research" to find useful tasks to tackle. However, this background work is strictly limited to read-only tools, meaning it cannot alter app data, send rogue messages, or hijack your browser. High-risk actions like updating passwords still mandate direct human execution, while an Activity View log lets you track everything it did while you were away. For enterprise clients, OpenAI is also collaborating with Microsoft to integrate specialist dots into Agent 365 governance controls.

Why it matters

We think dots signal the beginning of the post-chat interface era for personal AI.

For everyday pros and developers, this makes persistent agents accessible without the headache of self-hosting. In the open-source scene, tinkerers have tried to hack together similar workflows using tools like OpenClaw, but managing persistent compute, browser sessions, and auth tokens usually required serious technical patience. Bundling an isolated cloud machine directly into a standard consumer subscription removes almost all of that friction.

Across the industry, the battleground has clearly moved from raw model reasoning to persistent execution. With Meta developing Muse and xAI working on Grokbot, everyone is racing to become the background operating system for your daily work. OpenAI's move puts immense pressure on rivals to offer dedicated compute alongside their models.

That said, giving an AI continuous read-write access to over 4,000 services is a double-edged sword. Even with read-only background limits and auto-review safety filters, indirect prompt injection and silent logic errors remain real hazards once an agent acts on your behalf across Slack and email. It is a massive leap in convenience, but trusting it with consequential company workflows will still require careful oversight.