Free in six countries including Japan, for hundreds of millions of people
Google on Nov. 17 (Nov. 18 Japan time) announced that it has begun offering "Tasks," which lets users hand off longer projects and recurring schedules in the app for its AI chatbot "Gemini," free of charge. The countries covered are the U.S., Japan, Korea, Canada, Australia and India, and the official blog said it is rolling out to "hundreds of millions of people." Tasks runs on "Gemini 4 Hydrogen," the new model Google released in November.
The Gemini app's official account @GeminiApp announced it on X at 2:04 a.m. Nov. 18 Japan time. The post positions "Chat" for quick back-and-forth and "Tasks" for longer projects and recurring schedules, and describes those covered as "eligible users."
Put Gemini to work for you with Tasks. While Chat is for quick back-and-forth answers, Tasks are for longer projects and recurring schedules. Beginning today, Tasks are now free of charge for hundreds of millions of eligible users, starting in the U.S., JP, KR, CA, AU & IN. 🧵
The official blog was written by Josh Woodward, Google's vice president in charge of Labs, the Gemini app and AI Studio. Users switch the toggle on the app's home screen from Chat to Task and describe what they need, and Gemini works on it in the background.
Google has not disclosed the conditions for use in Japan, such as supported languages, age or account type, or any limit on how many times Tasks can be used. The blog and the thread also do not spell out who counts as an "eligible" user.
Hydrogen follows instructions and operates Google apps
@GeminiApp said that running Tasks on Gemini 4 Hydrogen makes it easier for Gemini to get things done, follow instructions more closely and call on the right tools for the job. Hydrogen is a model released on Nov. 12 (Nov. 13 Japan time); ORIENS covered its performance and pricing in an article at launch.
According to the thread, users can create and edit Google Docs and Sheets and draft and send Gmail messages inside Gemini without switching apps. They can also add multiple Google Calendar events at once with a single tap.
According to the blog, Gemini connects with more than 60 popular apps, and developers can apply to join the "Connected Apps Platform." As an example, Woodward described giving Gemini access to his electric utility's site through the new integration with "Google Password Manager" and having it analyze past billing statements in his email.
Phone calls on users' behalf start in the U.S., unavailable in five states
Google is starting a set of advanced features in the U.S. In the thread, it followed that announcement by introducing, in turn, phone calls on a user's behalf, checkout with Google Pay, email forwarding and requests by text. Google has not said when these features will be available in Japan.
With phone calls, Gemini calls businesses on the user's behalf. It can call several shops at once to compare quotes, check inventory or book an appointment, Google said. According to a footnote in the blog, calling is not currently available in five states: Indiana, Louisiana, Minnesota, Montana and Nebraska.
Google described safeguards for both the businesses receiving calls and users. Gemini identifies itself as an AI to the other party, and businesses can opt out at any time. Users can stop a call at any time while it is in progress and afterward review a summary and a full transcript in the chat.
On payment information, the thread said Gemini "never shares payment or sensitive account details over the phone." In the blog, Woodward wrote of his own case that Gemini won't give out his payment or private account information unless he says so.
We’re also rolling out a set of advanced features, starting in the U.S. Gemini can now place calls on your behalf, even dialing multiple shops at once to compare quotes, check inventory, or book an appointment. You can cancel anytime while a call is in progress, and review a crisp summary and full transcript right in chat. Gemini identifies itself, never shares payment or sensitive account details over the phone, and allows businesses to opt out at any time.
Google Pay checkout, with three card networks
Gemini can now check out with Google Pay using the "Agent Payments Protocol (AP2)," a framework for AI agents to make payments. Users set a spending limit and the stores it can use, and choose whether Gemini checks with them before buying or goes ahead on its own within the limit. Users still receive card benefits such as points, cashback and purchase protections, Google said.
Google is working with Visa and Mastercard, and soon with American Express, to secure purchases with digital tokens on supported cards. Receipts, order confirmations and delivery updates land in the payments app "Google Wallet." Google Wallet is expanding to iOS starting Nov. 17. Tasks in Gemini will soon integrate with the "Universal Commerce Protocol (UCP)" for shopping.
Google Wallet's connection to Gemini also gained a feature in November that answers personal finance questions (earlier article).
Requests by forwarded email and text, starting in the U.S.
According to the blog, starting in the U.S., users can forward emails to gemini@google.com or text Gemini to hand off tasks. Based on forwarded emails, Gemini can add events to a calendar or reorganize messy notes. Gemini replies in a private 1:1 thread with status updates and links to finished work, so there is never an accidental reply-all, Google said.
Texts can be sent via Google Messages or iMessage. According to a footnote in the blog, texting requires a U.S. phone number with an eligible mobile carrier.
Runs in a dedicated virtual machine and checks before buying
According to the blog, work assigned through Tasks runs on Google Cloud's "Agent Compute Engine" in a virtual machine dedicated to the user's account. The virtual machine is hardware-isolated and never shared with other users. It runs in an isolated environment (a sandbox) that Google's security red teams test regularly, the company said.
This is one of several layers of defense, which also include filters that detect malicious instructions hidden in search results and web pages (prompt injection) and "Google Safe Browsing," which limits which links Gemini can open. Gemini is designed to pause and check with the user before critical steps, asking for confirmation or pre-approval before clicking "Buy." Routine actions such as adding groceries to a cart can be set to "Always allow."
Google says mistakes remain, and some sites hand control back to users
Google acknowledged that while it tested Tasks extensively to minimize mistakes, it has not eliminated all of them. It said tasks may fail to complete or miss the mark, and asked users to report such cases with a "thumbs down."
Websites are also still adapting to AI acting on users' behalf, Google said. On some sites, Gemini makes it most of the way through and then hands the screen back to the user to click the final button. Google said it is partnering directly with major websites and security providers so that the web works well for both site owners and personal assistants.






