For example, the feature allows users to ask Gemini to find specific photos from past trips, gather information from recent emails, or pull together information scattered across different Google services. For instance, the assistant could retrieve details like license plate numbers from photos or plan a family vacation by analyzing past travel patterns and family interests stored across Gmail and Photos.

The new feature, which is being released in beta, will be rolled out on Wednesday to Google AI Pro and AI Ultra subscribers in the U.S., with a wider rollout to all users planned within the week. Once enabled, Personal Intelligence will work across Web, Android, and iOS and through all of the models in the Gemini model picker. The company also plans to eventually roll the feature out in Search’s AI Mode.

Google said in a blog post announcing the feature that the new capability “has two core strengths: reasoning across complex sources and retrieving specific details from, say, an email or photo to answer your question. It often combines these, working across text, photos, and video to provide uniquely tailored answers.”

The move is part of Google’s attempt to create an AI system that functions like a personal executive assistant, something industry leaders, including OpenAI CEO Sam Altman, Microsoft founder Bill Gates and Mustafa Suleyman, the DeepMind cofounder who now serves as CEO of Microsoft’s AI efforts, have long argued will be AI’s most valuable consumer use case. By giving Gemini access to a user’s Google ecosystem, the bot can automatically understand a user’s context and preferences without the need to specifically include all those details in the prompt.

Google already benefits from having extensive data on users—such as search history, emails, and calendars—which positions the company well to deliver this personalized experience with Gemini. If the feature gains significant adoption, it could strengthen Google’s competitive position.

All of the leading AI labs are currently working on AI systems that have some understanding of an individual user’s preferences and can handle tasks autonomously. But the data advantages that Google has through its ecosystem of popular products and apps may be difficult for others to match.

Microsoft has been expanding its Copilot platform with features including long-term memory, integration with services like Google Drive and Gmail, as well as introducing ‘Actions’ that allow the assistant to book tickets and make reservations on behalf of users. Anthropic also recently launched its Claude Cowork, a file-managing AI agent designed for non-technical users that works proactively within a user’s files.

Allowing AI agents this kind of deep access, however, presents data privacy and security issues. For example, in the case of Personal Intelligence, connected accounts could expose sensitive information like personal details from email correspondence, financial data from banking notifications, or location histories from photos and Maps activity. There’s also the risk of unauthorized access if a user’s Google account is compromised, potentially giving bad actors access to information aggregated from multiple services.

Google said the capability will be off by default to give users more control over privacy and security. Even when a user connects their apps to Gemini, the company said Personal Intelligence won’t be used for every response but only when Gemini determines that doing so will be helpful. Google is also asking users to give feedback about personalization via the thumbs-down button. The company is also introducing guardrails for sensitive topics, with Gemini trained to avoid making proactive assumptions about sensitive data, such as a user’s health details.