OpenAI has given ChatGPT a much closer view of what you are doing in your web browser. An update to the ChatGPT extension for Google Chrome, announced on 15 October 2026, lets the assistant read the tabs you already have open, sit in a sidebar with your full chat history, and, in its agent-style Work mode, carry out tasks inside those tabs such as filling in forms, sending messages and downloading files.
The change matters because it moves ChatGPT from a separate window you copy and paste into towards an assistant that works directly on the pages you are viewing, including sites where you are already signed in. That is convenient, but it also raises fresh questions about privacy and security. Here is what changed, how it works and what to check before you switch it on.
What OpenAI announced
According to The Verge, once the extension is installed, ChatGPT can view your browser tabs when you tag Chrome in ChatGPT's web or desktop app. OpenAI's demo showed a user asking ChatGPT to compare three open tabs of sofas they were considering. Engadget described the same example, with ChatGPT comparing the couches by dimensions, price, delivery time and return policy.
The update has three main parts:
- Tab awareness in Chat. In ordinary ChatGPT Chat, you can point the assistant at your open tabs and ask questions that draw on what is in them.
- A fuller sidebar. The ChatGPT panel inside Chrome now carries what OpenAI calls "the full power of ChatGPT", including your conversation history, so you can ask questions or hand over tasks without leaving the page.
- Browser control in Work. ChatGPT Work, OpenAI's mode for complex multi-step tasks, can now act in your Chrome tabs. The Verge lists sending messages, filling out forms, downloading files, and organising or grouping tabs among the supported actions.
The official ChatGPT account announced the sidebar upgrade on X on 15 October, and the extension's Chrome Web Store listing shows it was updated the same day.
Example tasks OpenAI suggests
OpenAI shared sample requests to show what the upgraded extension is meant for. As reported by The Verge, they include:
- Using the details in an event brief to fill out a registration form, stopping before submitting.
- Putting the vendors from several tabs into a comparison sheet, with pricing, key features and source links.
- Comparing contractor services across open tabs, then drafting an email to the best fit for a home renovation.
- Filling in a new-patient form for a dentist and adding the appointment to work and personal calendars.
Why OpenAI is leaning on Chrome
OpenAI used to offer its own web browser, ChatGPT Atlas. Engadget reports that the company retired that product on 9 August, and the Chrome extension has since become its main route into people's browsing. The Verge noted that the new browser-control feature is similar to Google's Gemini-powered "auto browse" feature in Chrome, which puts ChatGPT in direct competition with Google inside Google's own browser.
The extension is not limited to Chrome. Engadget, citing OpenAI, says it also works with other Chromium-based browsers, including Microsoft Edge, Brave and Opera. OpenAI's help documentation also lists Vivaldi. All five browsers support tab mentions and browser control from the ChatGPT desktop app, but the side chat panel is available only in Chrome, Edge, Brave and Vivaldi, not Opera.
Engadget reports that the extension is available on every ChatGPT plan. OpenAI's documentation adds that "browser availability can depend on rollout and your workspace settings", so not every user, or every company account, may see every feature straight away.
The update arrives during a busy month for AI assistants. For a wider look at how the leading systems compare, see our guide to the best AI models of October 2026 and our explainer on Google's Gemini 4 Argon launch.
How it works in practice
OpenAI's help pages describe a few ways to bring browser content into a conversation:
- Mention a tab. In the desktop app, and in the side chat where available, you can mention an open tab so ChatGPT uses that page as context.
- Highlight text. In browsers with side chat, you can select a passage and bring just that selection into the chat.
- Right-click. Choosing "Ask ChatGPT" from the page menu opens the side chat with the page already loaded as context.
For tasks that need ChatGPT to act on your behalf, the documentation says to install the extension through Settings > Computer Use in the ChatGPT desktop app, then start a Work or Codex chat and pick your browser with an @-mention. Chats started in the side panel also appear in the main ChatGPT app, so you can switch between the two.
The privacy and security questions
Letting an AI read and act on signed-in websites is powerful, and OpenAI's own documentation is frank about the risks.
What the extension can access
When you install it, Chrome asks you to approve a long list of permissions. According to OpenAI, these may include reading and changing all your data on all websites, reading and changing your browsing history on your signed-in devices, managing downloads, reading and changing bookmarks, and accessing the page debugger. OpenAI says these permissions are what let the extension run browser workflows, and that ChatGPT applies its own confirmations and settings on top of them.
The safeguards OpenAI describes
- Per-site prompts. By default, ChatGPT asks before it interacts with each new website. You can allow once, allow for that site, allow for all sites, or decline.
- Allowlists and blocklists. You can manage which domains ChatGPT may or may not use from the desktop app's settings.
- Browser history. OpenAI labels history access as "elevated risk". ChatGPT asks each time it wants to use it, and there is no always-allow option.
- Data storage. OpenAI says it does not keep a separate complete record of your browser actions. It stores browser activity only when that activity becomes part of the chat, such as text read from a page, screenshots or summaries.
OpenAI also warns users to "treat page content as untrusted context". That points to a known weakness of AI browser agents: a malicious web page can contain hidden instructions designed to trick the assistant. The documentation specifically warns that misleading page content could increase the risk of ChatGPT copying data from your history "somewhere unintended", and that "Allow for all sites" removes the confirmation step entirely. It advises against sending secrets or highly sensitive data through browser tasks unless you are present to review each prompt.
What it means for you
- Useful for research and admin. Comparing products across tabs, pulling details into a spreadsheet or drafting a reply based on several pages are the kinds of chores this update is designed for.
- Keep confirmations on. Use "Allow once" or per-site permissions rather than "Allow for all sites", especially if you use the same browser profile for banking, email or work tools.
- Be careful with money and identity. Avoid letting an AI agent handle payments, UPI apps, net banking or government ID portals. Our guide to protecting yourself from UPI and online payment frauds applies here too.
- Check before you submit. Ask ChatGPT to stop before sending a form or message, and review what it filled in.
- Consider a separate profile. Installing the extension in a dedicated browser profile limits what it can see.
- Office users: check policy. If you use ChatGPT through an employer, your workspace settings may limit these features, and company rules on AI tools still apply.
The bottom line
The upgraded extension turns your open tabs into working material for ChatGPT and lets it act for you inside the browser. That could save real time, but the same access widens what can go wrong, which is why OpenAI's own guidance leans so heavily on permissions, prompts and human review.
