Gemini Spark integrates with Chrome logs

On July 30, 2026, Google integrated Gemini Spark with Chrome, enabling the cloud agent to continue tasks in the user's browser, open sites with active sessions, and use saved login data with permission. The integration builds on Chrome Auto Browse, introduced on January 28, by bridging the cloud agent with the local Chrome browser. Gemini Spark runs on a remote Google machine and can perform long tasks even when the user's computer is off, but for accessing authorized sites, the local browser must remain on. Users see a plan before starting, must approve linking the device, and confirm each website access. Google Password Manager does not transfer passwords to the AI, and sensitive steps like payments are returned to the user. For managed browsers, Google added the policy GeminiSparkSettings, warning administrators of significant risks to credentials, local network access, and loss of contextual signals. The feature is currently available in the US for users aged 18+, with English device language, latest Chrome, Google AI Pro or Ultra subscription, standard or enhanced Safe Browsing, and no incognito mode. Pro allows up to 20 multi-step tasks per day, Ultra up to 200.