OpenAI has implemented a significant update to its Terms of Service, fundamentally changing how it utilizes user-generated content from its consumer-facing products. The new policy establishes that data submitted by users to services like ChatGPT and DALL-E will be used by default to help develop and improve OpenAI’s services, which includes training its artificial intelligence models.

This marks a shift from the previous policy. Formerly, user data was also used for training unless a user proactively completed a specific Google Form to opt out of data collection. The updated terms now make data usage for training the default setting for all consumer service users.

The New Opt-Out Mechanism

Under the new terms, OpenAI has introduced a different method for users to opt out. For ChatGPT, users can now prevent their data from being used for training by disabling chat history and training in the application’s settings. This provides an in-app toggle that was not previously available. However, the default behavior upon account creation is to have this data sharing enabled.

Distinction Between Consumer and API Data

It is critical to note that this policy change primarily affects OpenAI’s consumer services. The company’s policy regarding data submitted through its API and for its business customers, like those using OpenAI Business, remains unchanged. According to the terms, data sent via the API is not used to train or improve OpenAI models unless the user or business explicitly opts in to share their data for that purpose. This creates a clear distinction in data handling practices between individual consumer users and enterprise-level or developer clients.

Source: https://thezvi.substack.com/p/openai-moves-to-complete-potentially