LinkedIn Trains AI on User Data Before Updating Privacy Terms
LinkedIn scrapes user data to train generative AI models but only updates its terms of service after the fact. The platform currently offers an opt-out toggle for U.S. users, while European users remain exempt due to strict regional privacy laws.
LinkedIn trains generative AI models on user data without properly updating its privacy policy beforehand. U.S. users see an opt-out toggle in their settings that discloses the platform scrapes personal data for "content creation AI models," but this data use initially occurs without a refreshed terms of service agreement. European users in the EU, EEA, and Switzerland do not face this scraping due to strict regional data privacy regulations.
The company uses this collected data to train its own internal models for writing suggestions and post recommendations, but it also shares information with third-party providers like its corporate parent Microsoft. LinkedIn collects a wide range of personal data for this training, including posts, articles, usage frequency, and language preferences, while claiming to apply privacy-enhancing techniques to limit sensitive information in the datasets.
Users who wish to stop this data collection can navigate to the "Data Privacy" section in their desktop settings and toggle off the option for AI model training. However, LinkedIn explicitly notes that opting out does not reverse or affect any AI training that has already taken place using their data. Meanwhile, advocacy groups like the Open Rights Group push for regulatory investigations into LinkedIn and other social networks that engage in this default data harvesting.