As the use of ChatGPT continues to grow, questions about the privacy of users’ conversations are receiving renewed attention. Many people wonder whether their private chats can be seen by human employees or contractors and how their data is handled.
OpenAI says that ChatGPT conversations from consumer services may be used to improve model performance, depending on the user’s data-control settings. Users can turn off “Improve the model for everyone”, while Temporary Chat conversations are not used to train models.
At the same time, OpenAI acknowledges that human review is part of its safety and moderation processes. Its policies say authorized personnel may access certain user content in limited circumstances, such as investigating abuse or security incidents, providing support, handling legal matters, or improving model performance when permitted. Access is restricted through technical controls and confidentiality requirements.
OpenAI also says it uses privacy safeguards to reduce personal information in data used for model training, including systems designed to identify and mask personal information.
For users, the key takeaway is simple: ChatGPT should not be treated like a completely private diary. Avoid entering passwords, financial details, confidential business information or other highly sensitive data unless you understand the relevant privacy and data-control settings.
For particularly sensitive conversations, users can consider Temporary Chat and review their Data Controls to decide whether new conversations can be used to improve OpenAI’s models.
