opinion

Tumblr and WordPress to Sell Users' Data to Train AI Tools

Builders relying on third-party data for AI training must understand the provenance and consent terms, as using data from such programs could pose ethical and legal risks that affect project sustainability and reputation.

Hacker News AI·February 27, 2024·1 min readopinion

opinionTumblr and WordPress to Sell Users' Data to Train AI Tools

404media.co

What happened

According to Hacker News AI, Automattic, the parent company of Tumblr and WordPress.com, is planning to sell user data to AI companies for training models. The data will include public posts from both platforms, with users given the option to opt out but defaulting to inclusion. This move mirrors earlier decisions by platforms like Reddit and Stack Overflow, which have struck deals with AI firms amid growing demand for training data. Privacy advocates have criticized the lack of explicit consent and potential misuse of personal content. For developers and solopreneurs building AI workflows, this development underscores the importance of being aware of data sourcing practices when selecting training datasets, as well as the ethical and legal implications of using user-generated content without clear permission. It also highlights the need for transparency in how platforms handle user data in the age of AI.

Key takeaways

Automattic will sell public user content from Tumblr and WordPress.com to AI companies for model training.
Users can opt out, but the default setting allows data sharing without explicit consent.
The policy follows similar data-selling arrangements by other platforms like Reddit and Stack Overflow.
Privacy critics argue the lack of active opt-in undermines user trust and consent.
The move reflects the increasing commercialization of user-generated content as AI training material.