Google is automatically enrolling users in a new data-retention policy that feeds Search history—including images, voice recordings, and video—into AI model training for up to 4 years.
Meta paused its Model Compatibility Initiative following a security breach that exposed keystroke and screen-capture data to internal staff.
A security lapse left sensitive employee laptop activity accessible to Meta staff, raising questions about surveillance safeguards and consent.
Thousands of Meta engineers drafted into a data-labeling unit report morale collapse, with some calling the work 'soul-crushing' and describing their assignment as involuntary.
Google is splitting search data collection into a new 'Search Services History' setting, enabling AI training on Lens photos, Live recordings, and Translate audio—with an opt-out available.
Robotics firms are scaling physical AI training data by paying consumers and gig workers to record everyday tasks, raising privacy and labor questions.
A copyright case against Meta's Llama model now targets individual scientists, potentially shifting personal liability for AI training decisions.