You can now assign Gemini ‘tasks’ for free in six countries
Google released its Gemini tasks automation feature for free across six countries, letting users run Hydrogen-powered agent workflows without an AI Pro tier.
By Dillip Chowdary • Nov 18, 2026 • Source: 9to5Mac

At a glance
- Product
- Gemini tasks
- Previous name
- Spark
- Price
- Free
- Prior requirement
- AI Pro
- Model
- Hydrogen
- Available regions
- US, Japan, Korea, Canada, Australia, India
In this post
Google rolled out its autonomous tasks feature to free Gemini users across six countries on November 17, 2026, dropping the requirement for a paid AI Pro subscription, as detailed in 9to5Mac's report. The feature, originally developed under the name Spark and positioned by Google as a 24/7 personal agent, was brought into the primary Gemini interface during an initial rollout in October. The update allows users on standard free accounts to delegate multi-step automation workflows directly within the app rather than limiting them to conventional back-and-forth conversational prompts.
This deployment report covers how the interface switch works on web and mobile devices, the underlying architecture powering agent execution, and the security sandboxing applied to account data. It is aimed at developers, system administrators, and everyday users who manage personal information across Google accounts and need to understand the execution boundaries, geographical availability, and credential integrations involved in running automated agents on web platforms.
01Dedicated task selector in Gemini interfaces
Google placed the new functionality directly into the existing Gemini view, positioning a toggle between standard conversation and automated execution. Users accessing the mobile app see Chat and Task options positioned above the prompt box, while web users find the same controls directly below the main input area. Selecting Chat keeps users inside the standard conversational model, while choosing Task hands off autonomous operations to the background agent system.
The agent interacts directly with external web portals and internal account systems to parse complex files. In one real-world utility scenario provided by Google, the system analyzed a surprise 389-dollar electric bill by logging into a utility portal through Google Password Manager and scanning past billing statements found across connected email messages. The automated run identified that a heat pump running during 4 p.m. peak pricing hours caused the spike, extracting details from raw PDF attachments that manual review had missed.
02Shopping Graph and checkout integrations
The tasks system connects with Google Shopping Graph to parse commercial options, sort technical specifications, and automate purchases. During hardware selection tests, the agent compared multiple USB-C chargers, queried the user with follow-up questions concerning wattage and color preferences, and completed the checkout sequence using Google Pay. The agent also discovered and applied an active discount code that deducted 10 dollars from the total purchase price.
| Metric or Feature | Previous Requirement | Current Release |
|---|---|---|
| Account tier | AI Pro subscription | Free tier |
| Underlying model | Standard Gemini model | Hydrogen model |
| Interface mode | Chat only | Chat and Task toggle |
| Password storage access | Manual user entry | Google Password Manager |
| Payment execution | Manual checkout | Google Pay integration |
Google is positioning this deep ecosystem integration as an operational advantage for personal automation. The company noted that personal credentials, emails, calendars, passwords, photos, files, and payment cards already reside on Google infrastructure securely, which removes the need to transmit sensitive credentials and personal data to third-party agent services.
03Virtual machine sandboxing and prompt defenses
To prevent unauthorized remote actions, Gemini tasks execute inside dedicated hardware-isolated virtual machines. Each environment is strictly sandboxed to prevent unauthorized privilege escalation, while internal security filters provide continuous protection against prompt injection attacks. Network requests generated during automated navigation sessions route through Google Safe Browsing to evaluate external destinations before loading page assets.
Advertisement
Tech Pulse Daily
Get tomorrow's pulse first
Join engineers who read Tech Pulse before stand-up. Free, weekday mornings.
Google implemented human-in-the-loop controls to prevent accidental or malicious financial operations. The system requires explicit user confirmation before finalizing any financial transactions or payment submissions. When an incoming prompt lacks specific parameters or contains ambiguous instructions, the agent halts autonomous actions to ask clarifying questions before continuing.
04Terminal setup and model selection flags
Developers accessing automated Gemini agents through command-line utilities can update packages and target the Hydrogen engine using vendor switches. The environment must be updated through standard package registries before issuing agent execution flags.
npm install -g @google/gemini-cli/model hydrogengemini --model hydrogengemini config set default-model hydrogenNote: Free task automation is currently restricted to accounts located in the United States, Japan, Korea, Canada, Australia, and India.
05International deployment across six markets
The automated tasks release is powered by Google's Hydrogen model, which handles the planning steps and external data extraction required for long-running workflows. Google has deployed the free tier across six initial markets: the United States, Canada, Australia, India, Japan, and Korea. Users in these territories receive the interface toggle without paying for an upgraded tier.
The expansion marks a transition for consumer-facing AI agents from paid research experiments into standard utility tools. By combining native credential access, isolated virtual execution, and financial clearinghouse integrations into standard accounts, Google has made personal multi-step workflow automation accessible across six major international user bases.
Frequently asked questions
What was Gemini tasks called before this release?
The feature was previously known as Spark before Google integrated it directly into the main Gemini app in October.
Which countries have access to free Gemini tasks?
Free access is available to users in the United States, Japan, Korea, Canada, Australia, and India.
How does Google secure automated web actions in Gemini tasks?
Gemini tasks run inside hardware-isolated virtual machines with prompt injection defenses, Google Safe Browsing checks, and mandatory confirmation before purchases.
Sources
Author
Dillip Chowdary
Writes Tech Bytes coverage of AI, engineering, and the tools that actually ship. Editor of Tech Pulse Daily.
Related on Tech Bytes
Advertisement