Google’s New Gemini Spark AI Agent
As spotted by a user on X, Google appears to be readying a Gemini AI agent called Gemini Spark. Unlike other AI assistants that require prompts, the Gemini Spark is said to be designed to work continuously in the background instead of waiting for manual prompts.
Screenshots of the Gemini Spark onboarding process shared by the user show that Gemini Spark can access data from Connected Apps, skills, chats, tasks, websites where users are logged in, Personal Intelligence, location, and more. Gemini could also share certain information with third parties, including names, contact details, files, preferences, and other sensitive data.
The screenshot confirms that the Gemini Spark is an experimental feature. This AI agent is expected to ask permission before carrying out sensitive actions, but it could still perform actions such as sharing information or making purchases without asking in some situations. Users are advised to supervise the agent and avoid relying on it for medical, legal, financial, or other professional help.
Gemini Spark saves remote browser data, including login credentials and remote code execution. Google says users will be able to clear this data and turn off Connected Apps or other Personal Intelligence features in Settings. Activity related to Gemini can also reportedly be managed or deleted through Gemini Apps Activity controls.
Google is expected to unveil the Gemini Spark at I/O 2026. The annual developer conference is scheduled for May 19.
