OpenAI Announces New Ways To Manage Your Data in ChatGPT
- By Paul Mah
- May 02, 2023
OpenAI is giving users the ability to turn off the chat history in ChatGPT and ensure that the data fed to ChatGPT won’t be used to train and improve subsequent versions of the popular AI service.
This was announced in a blog last week, which highlighted the availability of the new option. This can be accessed from the “Settings” menu entry, by clicking on the “Chat History & Training” option.
In addition, a new Export feature will also allow users to export their ChatGPT data. Selecting the “Export data” option will also cause the ChatGPT service to send an email with a link to download a zip file containing all chat data.
By default, users consent to OpenAI using their data to improve the performance of its AI models. The only way is to opt-out by submitting a request using this form.
The ability to delete chat directly from the main web interface will no doubt be welcomed, though OpenAI warns that data will still be retained for a period of time to curb potential abuse.
“When chat history is disabled, we will retain new conversations for 30 days and review them only when needed to monitor for abuse, before permanently deleting,” noted OpenAI.
One key concern around the use of ChatGPT arose from potential privacy breaches as sensitive personal or corporate information might be inadvertently exposed. As we reported last month, Samsung Semiconductor saw at least three data breaches as top-secret source code and confidential test patterns were sent to ChatGPT.
To assuage such concerns, OpenAI is apparently working on a ChatGPT Business subscription for businesses who need more control over their data, as well as enterprises seeking to manage end users’ accounts from a central console.
ChatGPT Business will follow the data usage policies of its API service, which by default means that end users’ data won’t be used to train new AI models. ChatGPT Business will be made available “in the coming months”.
Image credit: iStockphoto/Oleksandr Shchus
Paul Mah
Paul Mah is the editor of DSAITrends, where he report on the latest developments in data science and AI. A former system administrator, programmer, and IT lecturer, he enjoys writing both code and prose.