OpenAI; Midjourney, Inc. × Automattic Inc.
Automattic, the parent company of Tumblr and WordPress.com, is providing user-generated content to AI firms OpenAI and Midjourney for model training. [32, 47] While talks were first reported in early 2024, an August 2026 report confirmed the collaboration was ongoing, with Automattic preparing to release an opt-out setting for users. [32] The price of the deal has not been disclosed.
The asset
A corpus of public user-generated posts from Tumblr and WordPress.com, reportedly spanning from 2014 to 2023, to be used for training generative AI models.
The record
- Seller status
- Going concern
- Sector
- Blogging platforms and web publishing
- Modalities
- text, images
- Jurisdiction
- US
- Personal data
- The seller, Automattic, stated it would provide a setting for users to opt out of having their data shared with third parties. [32, 46] Initial reports in 2024 alleged that an early compilation of the data mistakenly included private, deleted, and mature-rated posts. [46, 47]
What it means if you hold data
This deal demonstrates that large archives of diverse, unstructured user-generated content, including blogs and image posts, are a valuable asset class for training multimodal AI models. It also highlights the significant technical and policy challenges in sanitizing such data and respecting user privacy, as evidenced by the reported inclusion of non-public content and the subsequent need for opt-out mechanisms. [46]