OpenAI is testing an update to ChatGPT’s Temporary Chat feature that changes how personalization works in short-lived conversations. Temporary Chat is designed to function like a blank slate, allowing users to start sessions that are not tied to previous chats and do not add to long-term memory. The new test, however, introduces a middle ground between privacy and personalization.
Under the update, users would be able to keep their personal customizations active while still preventing the conversation itself from being saved or used to shape their broader account history. That means preferences such as tone, style, and certain memory-based settings could still apply, even though the chat session remains temporary.
The change was first spotted by AI engineer Tibor Blaho, who identified a hidden setting tied to the feature. According to his findings, the option allows personalization tools to function without permanently storing the conversation. The feature is not yet widely available and appears to be in an early testing phase.
Temporary Chat has often been compared to a browser’s incognito mode, offering a way to interact without long-term tracking inside the product. This update complicates that comparison by blending session privacy with retained user preferences. OpenAI indicates the feature will be optional and can be turned off.
Despite the temporary nature of these chats, the company may still retain copies for up to 30 days for security and abuse monitoring. The update reflects ongoing efforts to balance user control, personalization, and platform safety as AI tools become more integrated into daily workflows.