After harsh criticism: ChatGPT will no longer learn from your data

 


OpenAI is now giving users of ChatGPT more control over their data. The new feature allows users to deactivate the chat history, so that conversations are not used to train ChatGPT.

At the end of March, Italy became the first country to announce a ban on ChatGPT. The main reason cited was data protection concerns. There are also concerns in Germany regarding the rapid advances of artificial intelligence.

OpenAI, the company behind ChatGPT, is partially responding to this and offering users a new option to gain more control over their data.

ChatGPT and data protection: these are the facts you should know • Experts warn about ChatGPT: this is why you should use it with caution

ChatGPT: new option offers more data protection With a new option, users can now simply prevent ChatGPT from using the entered information to improve the language model. In other words, if desired, ChatGPT will no longer learn from your messages.

At the same time, with the new setting, the chat history will no longer be permanently saved. OpenAI will, however, store the history for 30 days and reserve the right to check messages for abuse.

The function is also not retroactive: your past chats will continue to be used to train ChatGPT.

The new option can be found in the settings under "Data Controls", somewhat hidden behind the "Show" button. If you slide the slider to "Off", your future chats will no longer be used to improve ChatGPT.

Although OpenAI stores the history for at least one month, users are denied access to previous chats. The sidebar with previous chats is still displayed but can no longer be clicked.

If you still want to access your previous chat history, you must either keep the learning function enabled or export your chat via the settings.

The data protection concerns regarding ChatGPT are by no means resolved, but the company seems to be paying more attention to the issue in the future.

With "ChatGPT Business", OpenAI has announced a special version of the language model for companies. This will no longer learn from user data by default.

Whether with or without a saved history: you should never disclose confidential information such as phone numbers or passwords. Researchers have already managed to extract such information from ChatGPT.

 

Post a Comment

Previous Post Next Post

Contact Form