Before you type: who can see your AI chats, and the setting that keeps them out of training

When using AI chat services, treat your interactions like postcards rather than personal diaries. Unless you adjust a specific setting, your typed words could potentially be used to enhance the AI algorithms and, in some cases, may be reviewed by human employees. Google's Gemini states that some reviewers, including those from their service providers, may examine collected data, with chat logs retained for up to three years even after user deletion.

Conversely, ChatGPT allows users to disable the "Improve the model for everyone" feature, preventing their new chats from contributing to model training while preserving access to old chat histories. The protection applies to sensitive information such as passwords, Social Security numbers, bank details, and medical records, as well as others' private information.

To enable this setting, users should navigate the account settings within the app: on a computer, access the settings menu, proceed to Data controls, and turn off "Improve the model for everyone"; on a phone, tap the profile icon in the sidebar, go to Data controls, and disable the same feature. Google mentions that without this setting, future chats won't be utilized for AI training unless users provide feedback, and the logs are kept for 72 hours.

Both platforms offer a temporary chat option, where ChatGPT chats are not used for training but may be retained for 30 days for safety measures, whereas Gemini's temporary chats are neither trained upon nor kept beyond 72 hours. However, providing thumbs-up or thumbs-down feedback could still lead to the inclusion of the conversation in training.

Lastly, corporate or educational accounts might have settings imposed by the respective organizations. To ensure privacy, users are advised to review and modify this single setting within their preferred AI chat application as soon as possible.