Google Launches Gemini: The AI That Exploits Your Personal Data

Google Unveils Gemini Personal Intelligence

Gmail, Photos, YouTube... What if your AI assistant truly understood your digital life? Google is testing Gemini Personal Intelligence, a feature that directly leverages your personal data. For several months, tech giants have been working to transform AI assistants into genuine co-pilots for everyday life. With Gemini, Google continues this strategy by introducing a feature that takes personalization to a new level: Personal Intelligence.

Launched in beta in the United States earlier this year, this feature is now beginning to roll out to other regions around the world, including India and several Middle Eastern countries. For Google, the goal is to enable Gemini to utilize users' personal data to provide more relevant and contextual responses.

An AI That Connects Directly to Gmail, Photos, or YouTube

With Personal Intelligence, Google is proposing a paradigm shift for its Gemini assistant, which was launched in 2023. Instead of being limited to information available on the web, the AI can now access data from the user's Google applications. Specifically, users can connect multiple services such as:

• Gmail
• Google Photos
• YouTube
• Google Search

Once these sources are linked, Gemini can analyze this information to answer very specific questions. In its official statement, Google provides the example of a user who might ask Gemini: "What are my travel plans for Jaipur?" The AI can then retrieve booking confirmations from Gmail, identify screenshots saved in Google Photos, or rely on recently viewed YouTube videos to suggest restaurants. This approach allows Gemini to cross-reference various types of content (text, photos, or videos) to reconstruct a precise context and generate a personalized response.

A Still Experimental Feature

However, Google acknowledges that this type of personalized AI is not perfect. The model can sometimes misinterpret the context of personal data or make erroneous connections between different elements. The company gives a telling example: if Gemini detects numerous photos taken on a golf course, it might conclude that the user enjoys this sport. However, these photos could simply correspond to moments spent with a loved one, unrelated to a personal passion. Similarly, the AI may struggle to understand certain nuances, especially when aspects of personal life evolve over time. Google encourages users to correct Gemini directly when an interpretation proves inaccurate.

Google Emphasizes Privacy Concerns

In light of concerns regarding the use of personal data, Google highlights several safeguards. The feature is disabled by default: users must manually choose which applications they wish to connect. They can also disable access or delete their history at any time. The company assures that content from Gmail or Google Photos is not directly used to train AI models. It is only used to respond to a specific query. The data utilized to improve the systems mainly concerns the prompts sent to Gemini and the generated responses, after anonymization.

For now, Personal Intelligence is reserved for paying subscribers of Google AI Plus, Pro, and Ultra plans. However, Google indicates that it intends to gradually extend the feature to free users in the coming weeks, as the company accelerates the rollout of Gemini across several strategic markets.