Google’s Disco browser shows how AI browsing trades context for constant observation
Disco’s GenTabs feature reveals a future where browsers must read everything on your screen to function, raising new questions about privacy at Chrome scale.
Google has introduced an experimental browser that treats the web as a single, readable workspace rather than a collection of tabs. The idea is simple. Instead of managing dozens of pages, users describe what they want to do, and the browser assembles tools to help them do it. That approach depends on context, and context requires visibility. For the system to function, Google’s AI needs to see what's open, what's being read, and how those pieces relate.
That experiment, called Disco, was announced by Google Labs on December 11, 2025. Built on Gemini 3 and powered by a feature known as GenTabs, Disco goes further than existing browser assistants. Rather than assisting within pages, it generates new interfaces based on what a user is researching. Planning, comparison, and organisation are no longer manual steps. They're inferred and constructed in real time.
This works best when browsing becomes fragmented. Multiple tabs for travel planning, product research, or study can be collapsed into a single interface that draws from everything already open. The browser effectively becomes a temporary workspace, assembling timelines, comparisons, and reference material without requiring the user to organise them first.
Trip Planner Demo (Credit: Google)
That capability has a technical cost. GenTabs relies on broad access to on-screen content, including open tabs and active pages, and uses chat history to infer intent. There are no selective permissions. The system is either allowed to observe everything in the browser or it cannot operate. This isn't an implementation detail. It's the core design choice.
What appears on a screen is often more revealing than explicit inputs. Research into health conditions, financial activity, internal documents, and private correspondence all occupy the same visual layer. Disco is designed to remove friction between thought and action, and in doing so, it treats visibility as a prerequisite rather than a risk.
This design arrives at a moment when the privacy implications of AI browser tools are already under scrutiny. In August 2025, researchers at University College London analysed several popular AI browser assistants and found that many transmitted sensitive webpage content and user inputs to external servers. In some cases, data handling practices conflicted with existing protection standards. The issue was not isolated misuse. It was a consequence of how these tools function.
Loy Okezie
Google has stated that Disco operates with user consent, but when continuous observation is required for basic functionality, consent becomes a limited safeguard. Users can agree to access, but they cannot meaningfully constrain it without disabling the system entirely.
Disco also reflects a broader shift rather than a standalone experiment. Perplexity, OpenAI, and Opera are all developing AI-native browsers that prioritise autonomy and context awareness. What distinguishes Google’s approach is proximity to Chrome. Disco exists as a testing environment, and Google has indicated that successful ideas may later influence its mainstream products.
If those features move into Chrome, they wouldn't be adopted gradually. They would become defaults. With more than 3 billion users, changes to how Chrome reads and responds to browsing behaviour would redefine expectations across the web.
GenTabs Workflow (Credit: Chrome)
Google has now opened a waitlist for macOS users in the United States. The sign-up is a simple Google Form that doesn't even require a Google account, asking only about your previous AI tool usage. The company has been transparent about looking for testers willing to push the limits and help define where this technology goes next.
Overall, the practical question is no longer whether AI browsers are useful, but how they should be used. One emerging response is separation. AI-driven browsers can function as public workspaces for research, planning, and synthesis, while traditional browsers remain reserved for financial activity, health information, and private communication.
The shift toward AI-mediated browsing is already underway. What remains unsettled is whether users will establish boundaries before observation becomes a standard requirement of using the web.
