Google made Tasks in Gemini free for hundreds of millions of people across the United States, Japan, Korea, Canada, Australia and India. The agent runs on the new Gemini 4 Hydrogen model and can now call businesses, pay bills and finish errands on a user’s behalf.
What Tasks Can Do Now
Tasks started as Gemini Spark, a paid feature for AI Pro and Ultra subscribers. Users flip a toggle on the Gemini home screen from Chat to Task, describe the job, and the agent works in the background, sending a notification when the result is ready.
The free rollout runs on Gemini 4 Hydrogen, the model Google introduced for fast agentic work. Tasks can dig through email and files, compare products with Google’s Shopping Graph, and check out with Google Pay, finding discount codes along the way.
Calls, Payments and New Channels
The advanced features start in the United States. Gemini can call local businesses, identify itself as an AI at once, and sit through hold music. Businesses can opt out, and Gemini will not share payment or private account details unless the user allows it.
Purchases run through the Agent Payments Protocol, with digital tokens from Mastercard and Visa, and American Express coming soon. Users set a hard spending cap, keep their card points and protections, and receipts land in Google Wallet, which is also expanding to iOS today.
Google is also opening new ways to reach Gemini. Users in the United States can forward emails to a dedicated Gemini address or text the assistant through RCS, and it can work with more than 60 connected apps, from calendars to shopping services.
The Safety Net
Every task runs inside a hardware-isolated virtual machine that belongs to one account and is never shared. Prompt-injection filters scan the pages Gemini opens, Safe Browsing limits which links it can follow, and the agent pauses to confirm before critical steps such as buying.
Google is candid that the agent still makes mistakes. A task can fail or miss its objective, and on some websites Gemini gets most of the way there, then hands the screen back so the user clicks the final button. Feedback through a thumbs-down helps Google improve it.
Our Read
This is Google’s clearest move yet from chatbot to agent. Making Tasks free puts a working agent in front of hundreds of millions of people who already live inside Gmail, Calendar and Google Pay, which is exactly the data advantage rivals cannot copy quickly.
The real test is trust. An agent that calls shops and spends money needs to be right far more often than a chatbot that merely answers. Google’s spending caps, confirmations and isolated machines are the right scaffolding, but one high-profile billing mistake would set the whole idea back.
The timing matters too. OpenAI already lets its agents pay through ChatGPT Wallet, and the two systems now compete directly on who can be trusted with a wallet. Google’s edge is distribution: Tasks ships inside an app hundreds of millions already use.
Why It Matters
Agents that can act, not just answer, are the next contest in AI. Google has now made that capability free at a scale no rival matches, in six countries at once. If Tasks works reliably, the assistant stops being a search box and becomes the person who runs your errands.
Background
Tasks began life as Gemini Spark, announced in May as Google’s always-on personal agent for paying subscribers. The free rollout follows Hydrogen’s debut in search last week and Gemini’s move into hospitals, as Google pushes the assistant into daily life.
Source: Google blog and Unite.AI report.


0 Comments