During the I/O 2024 event, Google shared numerous updates for its AI, Gemini. The company introduced the Gemini 1.5 Pro and Flash models, which garnered significant attention for their ability to generate realistic conversations. However, closer examination of Google’s data retention policies revealed privacy violations.
Google is recording chat histories to improve the Gemini model. Users are provided with an option to disable the history, but it has been revealed that Google stores certain conversation data for up to three years even when this setting is disabled.
The recorded data includes device and account information, language preferences, and user location data. The tech giant believes continuous data collection is necessary to train and improve the language model. However, experts argue that this poses a privacy risk.
One of the biggest concerns stems from the sharing of personal details during conversations. Gemini could potentially reveal personal details in its responses in other chats using this data, leading to potential data leaks.
Moreover, using data from users with different worldviews poses significant risks. Despite Google’s sensitivity to this issue, Gemini could gradually begin to support discriminatory views over time.
Google acknowledges that it stores user data in a controlled manner and uses it in the training of the model. However, the vast repository of personal information it collects also increases the risk of data breaches.
Page Contents
Toggle