Data privacy for latest features (GPTs and Assistants)

Hi all! Curious what the exact data privacy policies are for the new GPTs and Assistant APIs. Will OpenAI be able to keep the data used to create GPTs/assistants or the content created via interaction with them?

Up until you delete the data

OpenAI will retain language data for up to 30 days for purposes such as enforcing policy violations.

API model use is not used for training.

1 Like

In your gpts, you can uncheck the box “Use conversation data in your GPT to improve our models” to prevent OpenAI from using your conversations to train its models

Well hidden option :wink:


I’m not quite sure if using data to train models refers to training the GPTs model here with me, which would only affect this GPT, or if it refers to training the underlying general model of OpenAI.

1 Like

It refers to adding conversational data such as user inputs and quality of AI replies to datasets that might be used for pretraining or fine-tuning other AI models in the future.

I came here looking for privacy policy updates around public-facing, website chatbots powered by OpenAI GPT Assistants - which are a slightly different flavour than ones used for internal use (aka Custom GPTs etc.) Any direction on this related topic?

1 Like

The entry point to all OpenAI Terms and Policies.

The bulk of API is under business terms and usage policies.

ChatGPT GPTs and actions are covered under separate consumer terms, so can’t be directly compared.

In GPTs Data Privacy FAQs is mentioned but I can’t find it.

It looks like they removed the checkbox, which allowed you to uncheck the data training option. Does Anyone know if they simply require you to allow the use of input and output now or if they decided to change the default to 'can not train?"

I have found the solution, the above option appears only after I have sent at least one file for learning.