Gemini Spark agent reaches AI Pro in 160+ countries
Google opened its background-running personal agent to AI Pro subscribers in more than 160 additional markets, a week after the tier launched in the US.
Google has begun rolling out Gemini Spark, its persistent personal agent, to Google AI Pro subscribers outside the United States, extending availability to more than 160 additional markets. The company announced the expansion through its Gemini app account on Friday and published an updated availability page alongside it.
Spark had previously been restricted to the higher-priced AI Ultra tier internationally, with AI Pro access limited to the US for roughly a week. The distinction matters because Spark is not a chat feature: it operates a remote browser, calls connected third-party apps, watches for conditions to change, and continues executing after the user closes the conversation. Moving it down to the $20-a-month tier across most of the world is the largest single expansion of a consumer agent of this kind to date.
Google has paired the rollout with constraints. Sensitive steps — online payments in particular — require explicit human approval before the agent proceeds, and the company says it has added defences against prompt-injection attacks, the principal attack surface for any agent that reads untrusted web pages while holding a logged-in session. The Chrome auto-browse capability that lets Spark use saved credentials for errands such as booking apartment viewings or comparing flights remains a US-only feature for now.
Several jurisdictions are excluded from the expansion, and Google has not committed to dates for them. In practice the omissions track regions where agent behaviour intersects with unresolved regulatory questions about automated decision-making and data handling, leaving European users on the Pro tier without a product now available across much of Asia, Latin America and Africa.
Why it matters
Agents that act with a user's credentials have so far been sold as premium, small-audience products, which has kept their failure modes contained. Pricing one into a mainstream subscription tier across 160-plus markets converts prompt injection and mistaken autonomous actions from a research topic into a consumer-support and liability problem at scale — and sets the reference price that rival assistant makers now have to answer.