Gemini has learned to analyze personal data to personalize responses.

Gemini can now access data from Gmail, Google Photos, search, and other products to find and match relevant information without explicit user input. For example, when asked to buy car tires, the assistant can determine the right size by analyzing photos from Google Photos. It can also link email threads to YouTube videos watched or extract specific data, such as the user's vehicle information.

Google specifically emphasizes that user privacy is the central principle of this feature. App connectivity is disabled by default, and users can choose which services to connect to Gemini and can disable the feature at any time. Data from Gmail and Google Photos is not used to directly train the AI ​​models. It is used solely to generate responses to specific queries. Training occurs on anonymized data, such as the user's queries and the model's responses, so the system understands how to find information without storing personal data.

To ensure transparency, Gemini will endeavor to cite the sources it relies on in its response. Users can correct inaccuracies, request a response without revealing personal information, or use a temporary chat.

Personal Intelligence is currently in beta mode and is available to Google AI Pro and Google AI Ultra subscribers in the US.