SIAgentNews.com
← All news
DevelopingGeminiAgents

Google makes Gemini Tasks agent free for millions of users

Google is opening up its advanced agentic AI to free users in six countries, allowing the tool to perform multi-step actions like shopping and calling businesses.

· First reported by Android Central

6Countries where Tasks is free

The short version

  • Tasks was previously restricted to paid AI Pro and AI Ultra tiers.
  • The agent can perform actions like making phone calls, shopping, and managing emails.
  • Users can switch between 'Chat' and 'Task' modes to delegate complex workflows.
  • The tool runs on the new Gemini 4 Hydrogen model.
  • Security measures include isolated virtual machines to protect against prompt injection.

Why it matters

This shift moves Gemini from a conversational chatbot to a functional personal assistant that can execute real-world actions. For the average user, it means the AI can now handle logistics like trip planning or business calls autonomously.

From Chat to Action

Google is significantly expanding the capabilities of its Gemini AI by making its 'Tasks' agent free for a large portion of its global user base. This feature, which was previously only available to those paying for AI Pro or AI Ultra tiers, is now rolling out to users in the US, Japan, South Korea, Canada, Australia, and India. As Android Central reported, the move aims to provide millions of people with a more functional AI experience.

The Tasks agent represents a shift toward 'agentic' behavior, where the AI does more than just provide information. Instead of just answering a question, the tool is designed to run multi-step requests in the background. This means you can give the AI a complex goal and then go about your day while it works through the necessary steps to complete it.

The transition is powered by Google’s new Gemini 4 Hydrogen model. This underlying technology allows the AI to handle deeper integrations across various apps. Users can now switch between a standard 'Chat' mode for conversation and a 'Task' mode for delegating work that requires multiple actions or external interactions.

Real-world capabilities

The new free access brings several practical capabilities to the Gemini ecosystem. The agent can now perform actions such as calling local businesses using Google Duplex and making purchases through Google Pay. It can also receive text messages via RCS and handle forwarded emails directly through Gmail. These features allow the AI to act as a coordinator for your daily logistics.

Google is also deepening how Gemini interacts with the Workspace suite. The AI can now more reliably draft or send emails and create or edit documents in Google Docs and Sheets. It can even add multiple events to a Google Calendar without requiring the user to toggle between different applications. This creates a more seamless flow for people who use Google's productivity tools for work.

To access these features, users can flip a toggle located above the mobile prompt box or below the web prompt box. This allows them to switch from a simple conversation to a task-oriented workflow. You can also bypass the main app by forwarding emails or texting Gemini directly through Google Messages and iMessage, making the assistant more accessible across different platforms.

Security and privacy

Because the Tasks agent can access sensitive information like passwords and payment methods, Google has implemented specific security measures. When the agent browses the web to complete a task, it runs inside an isolated virtual machine. This environment is designed to protect the user against prompt injection, which is a type of attack where an AI is tricked into performing unintended actions.

Google also uses Safe Browsing protections to ensure the agent interacts with reputable websites. To prevent unauthorized spending, Gemini is programmed to ask for user confirmation before it completes any purchase. If the instructions provided by the user are not clear enough for the AI to proceed safely, the agent will ask for clarification before taking action.

The system is designed to keep credentials secure by using the Google Password Manager. This allows the AI to log into websites without ever exposing the actual passwords outside of Google’s own ecosystem. These layers of protection are intended to give users more confidence when allowing an AI to manage their digital lives and personal finances.

Limitations and next steps

While the release of Gemini Tasks is a major step forward, the technology is not yet perfect. Some websites may still prevent the agent from completing a specific task, which will require the user to step in and finish the final steps manually. This indicates that while the AI can handle much of the heavy lifting, it still requires human oversight for certain complex web interactions.

Google is continuing to expand the capabilities of the agent, with some new features initially launching only in the US. The goal is to move closer to a true personal assistant that can handle recurring tasks and research topics autonomously. As the model matures, the goal is to reduce the amount of 'babysitting' a user has to do for every individual step of a request.

For now, the tool serves as a bridge between a simple chatbot and a fully autonomous agent. It offers a way to manage a morning briefing or plan a trip while the user is busy doing other things. As Google continues to refine the Gemini 4 Hydrogen model, these capabilities will likely become more reliable and less prone to hitting roadblocks on third-party sites.

What to watch

  • How many more countries will receive the free Tasks agent?
  • Which specific websites will remain blocked to the Gemini agent?
  • How often will the agent require manual intervention for multi-step tasks?

Try it

Look for the toggle above your mobile prompt box to switch from 'Chat' to 'Task' mode.

Sources and links

First reported by Android Central. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

More AI news