Gemini Spark expands to 160+ countries with Chrome auto-browse

The country count is the headline, but the shape of what Google is now trusting a consumer agent to do is the more interesting part. Gemini Spark, the personal agent Google launched in May, is being wired into Chrome's auto browse feature, which lets the assistant operate against logged-in accounts and saved passwords on the desktop and do things like scheduling viewings for apartments you've saved or researching flight options, according to 9to5Google.

That is a real step up in what a mass-market subscription agent is being trusted with. Handing a chatbot a research task is one thing; handing it your session cookies and stored payment details is another. Google's stated guardrails are that Spark asks users to approve payments and other sensitive actions, and that the browser tool ships with protection against prompt injection. The Chrome piece is rolling out first in the US with more regions coming, while Spark itself now reaches over 160 additional countries for AI Pro subscribers, joining tools it already had including Connected Apps for Workspace and Search, Personal Intelligence, a remote computer with code execution, Canvas, and MCP integrations.

The honest caveat is that protection against prompt injection is a claim, not a measurement, and the reporting does not detail how the auto-browse surface logs actions, how the approval prompts are worded, or how password access is scoped. Take those specifics as reported, not settled. It also does not spell out which of the 160-plus countries are covered, or whether availability tracks language or Workspace tenancy.

For AI Pro subscribers outside the US who have watched agentic-browser demos from a distance, this is the moment the product actually turns up on their machine. For anyone building on top of Gemini through MCP or Connected Apps, the addressable surface just widened, because the assistant on the other end can now click through a checkout rather than merely describe how to. The question worth watching next is whether users trust the approval step enough to keep clicking through it.