LinkedIn will soon train AI models with data from European users (hostvix.com)

LinkedIn will begin using EU/EEA, U.K., and Swiss members’ data to train generative-AI models on November 3, 2025, citing GDPR’s “legitimate interests” as its legal basis and offering an opt-out labeled “Data for Generative AI Improvement.” Data types called out include profile details, public posts, articles, comments, saved résumés and job-application answers; private messages are explicitly excluded and LinkedIn says it screens out users it believes are secondary-school minors. Turning the toggle off stops future use of your data for training but does not remove information already used to build models. The move reverses a 2024 pause following criticism and limited legal challenges, and LinkedIn says AI feature access won’t require opting in. For AI/ML practitioners and privacy specialists this matters on two fronts: technically, LinkedIn will supply domain-specific training signals (professional profiles, recruiter Q&A, public career content) that could materially improve content generation and candidate-matching models; legally and ethically, the company’s use of “legitimate interests” (with an objection mechanism rather than affirmative consent) sets a notable precedent for how platforms may lawfully incorporate user data into ML pipelines in regulated markets. There’s also a parallel, separate expansion of data sharing with Microsoft for ad personalization in some regions—governed by different controls—so expect region-specific labels, opt-outs and potential dataset bias if many users exercise the objection.