Gemini Spark adds Chrome integration and expands to 160 more countries

Google has added direct Chrome integration to Gemini Spark, its autonomous personal AI agent, letting the agent use a person's logged-in accounts and saved passwords to complete web-based errands on its own. The company also expanded access for Google AI Pro subscribers to more than 160 additional countries.

What's new

According to Google's official announcement, "Now, with your permission, Spark can use your logged-in accounts and saved passwords to handle tedious web errands." That means Spark can now carry out multi-step browser tasks -- Google cites examples like scheduling apartment viewings or researching and booking flights -- without a person manually clicking through each step.

Because the feature hands an AI agent access to authenticated sessions, Google says it built in specific guardrails: "This is all done while protecting against threats like prompt injection and keeping you in the loop on sensitive actions." Sensitive actions such as payments still require explicit user involvement rather than running fully autonomously.

On availability, Google says the rollout starts in the U.S. and widens access for subscribers elsewhere: "We're rolling out access for Google AI Pro subscribers in over 160 additional countries starting today." Further international expansion is planned beyond this initial wave.

Context

Gemini Spark launched as a 24/7 autonomous personal AI agent, with Google first rolling it out broadly in India in late July 2026. That initial release focused on always-on task handling; this update is the first major capability expansion since then, moving Spark from a general-purpose assistant into a tool that can act directly inside the browser using a person's own accounts.

The move puts Spark in more direct competition with other browser-native agents that have emerged across the industry this year, as labs race to let AI assistants take real actions on the open web rather than just answering questions about it. Prompt injection -- where malicious instructions embedded in a webpage try to hijack an agent's behavior -- has become the central safety concern for this category of product, which is why Google is calling out its defenses explicitly in the announcement rather than treating them as an implementation detail.

Why it matters

Granting an AI agent access to logged-in sessions and saved passwords is a meaningful trust escalation compared to a chatbot that only reads and summarizes information. It shifts Spark from advising a user toward acting on their behalf inside real accounts -- email, travel sites, service portals -- which raises the stakes if the agent misinterprets a task or is manipulated by a malicious page.

The scale of the expansion is also notable: extending Google AI Pro access to more than 160 additional countries signals Google intends Spark to be a mainstream, globally available feature rather than a narrow regional pilot. Combined with the Chrome integration, this positions Gemini Spark as one of the more aggressive bets among major labs on agents that operate directly inside a user's existing web accounts, ahead of rivals still limiting agentic browser access to sandboxed or read-only modes.