LinkedIn will use member data, profiles, and public posts to train AI models (www.linkedin.com)

LinkedIn (Microsoft-owned) updated its terms: starting November 3, it will use member data — profiles, public posts and other non-private content — to train its generative AI models. Private messages are excluded, but the default setting is opt‑in; EU users are told LinkedIn relies on “legitimate interest” as the legal basis. Users can opt out, but must actively change the “Data for Generative AI Improvement” setting in LinkedIn’s privacy controls. This matters because LinkedIn’s corpus is uniquely rich in professional, industry-specific language and signals, which can materially improve model performance for business, hiring and content-generation tasks. At the same time it raises consent, IP, provenance and bias concerns: using user-generated professional content by default creates a platform feedback loop (platform content trains models that then produce more platform content), complicates attribution and redress, and may surface proprietary or sensitive information in model outputs. For practitioners and organizations, the change is a reminder to review data governance, opt‑out choices where appropriate, and consider downstream risks when models are trained on user-contributed professional text.