Google recently launched a new feature in Gemini that is deeply bound to Google Photos, allowing this AI assistant to extract information directly from the user's photo library to generate more personalized image content.Google's goal is for systems to "use real images of you and your loved ones" when creating AI images, rather than relying solely on users providing lengthy descriptions of appearance, family members and personal style in prompts.
The change is part of Google's larger "Personal Intelligence" strategy, which aims to connect Gemini with apps such as Photos, Gmail, Search, Maps and YouTube to incorporate more personal context into answers and make them appear more relevant to users. In terms of image generation, Gemini will use photo information obtained by connecting to Google Photos to automatically fill in details, trying to reduce the trouble of writing "mini-novel" long prompt words in traditional prompt projects.
The underlying image model driving this feature is Nano Banana 2, which is Google’s recently upgraded image generation model. Officially, it is positioned as a way to more quickly generate AI scenes that “look really relevant to you.” In short, if the user agrees to turn on this feature, Gemini will automatically refer to relevant photos in the user's album when generating pictures to adjust the character's appearance, scene style and other elements to make the result closer to you and your family in real life.
However, as soon as the function was exposed, it triggered a new round of privacy alarms and public controversy. Many critics worry that family photos, personal memories and even various sensitive photos will be further involved in the "AI content factory" and converted into part of machine-generated content, reinforcing the fear of "personal data being reprocessed into AI production materials." In the context of Microsoft's Recall feature, which has received strong backlash due to privacy issues, this type of AI innovation that "goes deep into end-user private domain data" is particularly sensitive.
Judging from the information disclosed by Google, this is not a service that is turned on by default, but uses a voluntary "opt-in" mechanism. Google said that "personal intelligence" is an optional feature, and users can decide which applications they want to connect to Gemini. The first batch of objects to access the Photos image generation function are eligible Google AI subscribers in the United States. In other words, Google currently does not directly scan and analyze all Google Photos albums without the user's knowledge.
In terms of data usage, Google emphasizes that the Gemini application will not directly use the user's private Google Photos library for model training. At the same time, the user can click the "Sources" button after generating the image to see which photo was selected by the system as a reference to guide the generation process of the current results. These instructions attempt to alleviate some of the concerns of outsiders and provide users with a level of visibility and traceability.
However, these assurances did not completely dispel doubts. Privacy advocates point out that the credibility of so-called "optional" systems depends largely on how clear and honest companies are in describing their features, and Big Tech's track record in this regard is not reassuring. Even if not directly used for training, the fact that users' private images are used for real-time inference and content generation may still be disturbing to many people.
Generally speaking, the current situation is not that Google is secretly "turning it on with one click" and scanning all users' photo libraries by default. But "letting AI roam freely in your private photo album and extract memory fragments to generate new content" itself is a concept that can easily make ordinary users feel defensive. At a time when the boundaries between AI services and personal privacy are constantly being redrawn, such product attempts around "personal intelligence" are likely to continue to become the focus of technological innovation and privacy disputes for some time to come.