Asger Risom
Google is testing a new Gemini feature that allows the AI assistant to autonomously book taxis or reorder food on a user’s phone, while still letting the user monitor and stop the process.
Google is testing a new version of its Gemini assistant that can autonomously complete tasks on your smartphone — including ordering food or booking a ride — without constant user input.
The feature turns Gemini into a mobile AI agent capable of handling multi-step actions in the background, while still allowing users to monitor or interrupt the process at any time.
How Gemini’s autonomous mode works
According to Google, users will be able to delegate specific tasks to Gemini, such as booking a taxi home or reordering groceries. Instead of simply suggesting options, the assistant can carry out the necessary steps itself.
The system is designed to work in the background, meaning it won’t interrupt the user while completing the task. However, notifications will allow users to follow the assistant’s progress, take control if needed, or cancel the operation entirely.
Google says the feature has been built with security limitations in place. Gemini will only have access to a restricted set of apps rather than the entire phone. When performing a task, the relevant app runs inside a secure virtual window on the device.
Limited rollout for now
The autonomous assistant feature is still in testing and currently supports only a limited range of applications, mainly in food delivery, grocery shopping, and ride-hailing services.
For now, it is available only in the United States and South Korea, and only on Samsung Galaxy S26 and Google Pixel 10 series devices.
If widely adopted, the feature could mark a shift from AI assistants that merely respond to prompts toward systems that actively execute real-world actions on behalf of users — raising both convenience and oversight questions.
Source: elEconomista.es
