AI Signal 151
OpenAI turns on model training by default for consumer plans as contractors review anonymized chats
↗
Internal OpenAI documents reveal that contractors evaluate anonymized user prompts and chats to improve models, and model training is enabled by default for consumer plans, meaning user conversations may be read by humans.
For engineers and users, this means ChatGPT consumer conversations are not private by default; they may be used for training and reviewed by contractors. This raises consent and data-handling concerns, especially for sensitive information shared in chats. Understanding the default settings is crucial for anyone building on or using OpenAI's consumer products.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Model training is on by default for consumer plans, so user prompts are used for training without explicit opt-in.
Contractors evaluate anonymized prompts and chats to improve the models, meaning human review is part of the process.
The chats can include sensitive personal information, raising privacy implications for users.
THE CLUSTER