ChatGPT can now read and act in your open Chrome tabs

OpenAI added two capabilities to ChatGPT's Chrome extension, announced on 15 or 16 October depending on the source. In Chat, ChatGPT can look across open tabs to gather information. In Work, its mode for longer tasks, it can control Chrome tabs to send messages, fill forms, download files and organise tabs.

OpenAI's example is comparing product listings across open pages without copying content. In Work, filling a booking form or downloading a report from a dashboard can be delegated. Actions happen on sites where the user is already signed in, so the agent works with the user's identity and permissions.

Android Authority reports the features are on all plans including free; PPC Land notes sources are unclear on whether tab control works on every plan.

OpenAI asks users to treat page content as untrusted context and review the site before letting ChatGPT continue, because of prompt injection: instructions hidden on a page could steer the agent on another signed-in site. The same week, a large AI-agent hacking campaign came to light.

Competition has moved inside the browser. Google built Gemini into Chrome; OpenAI, after closing its Atlas browser, moved into the browser people already use. Not every website visitor is human any more.

Browser agents have multiplied over the past year. Google built Gemini into Chrome with auto-browse features, including on Android. Anthropic's Claude can act in tabs through its own browser extension, and Perplexity offers its Comet browser. OpenAI chose to drop its separate browser effort and focus on Chrome, which most people already use, a faster route to distribution.

• Run the agent in a separate browser profile, away from banking and admin panels.
• Make forms agent-friendly with clear labels and error messages.
• Add a second confirmation step for payments and account changes.