Tinder, a US-based dating application owned by Match Group, is testing a new artificial intelligence feature that analyzes users’ camera rolls to generate insights aimed at improving profile recommendations and matches.
The feature, called “Photo Insights,” requires users to opt in before it can access images stored on their devices. Once enabled, the system scans photos locally to identify patterns related to interests, lifestyle, and personality traits. These insights are then used to suggest profile pictures and tailor matchmaking results.
According to the company, selected images may be temporarily uploaded to its servers and shared with a third-party AI provider to generate these insights. Tinder said the uploaded photos are deleted unless users choose to add them to their profiles, and unused data is scheduled for removal within a defined period.
The process used to select which images are uploaded has not been fully detailed. The company said the system evaluates photos based on factors such as quality, content, and similarity when determining which images are relevant for analysis.
Security researchers said the feature introduces potential risks related to data handling and transparency. They noted that users cannot independently verify how long images remain stored, whether deletion processes are enforced, or how third-party providers manage the data.
Researchers also raised concerns about the possibility of linking images to user identities if additional data is shared alongside photos. They said that if such datasets were exposed or misused, it could increase the risk of sensitive personal information being accessed.
Tinder said the feature is optional and can be disabled by users who do not want their camera roll analysed. The company stated that the tool is designed to improve personalisation and help users select images that better represent them on the platform.
The feature is part of a broader set of AI-driven updates aimed at refining how users are matched, including systems that analyse behaviour, preferences, and visual data to influence recommendations.