Tumblr's parent company, Automattic, is in talks with OpenAI and Midjourney to sell user posts for training AI models. Automattic is preparing to release a setting that allows users to opt out of data sharing with third parties. They have already scraped all public posts from Tumblr between 2014 and 2023, although there were errors including some non-public posts, and it remains unclear how these data will be handled and which data will be used for training models. Previously, Reddit had an agreement with Google to use user data annually for training Google's AI models; meanwhile, Shutterstock signed an agreement with OpenAI to train models using its photo library.