- New Feature: OpenAI is testing an upgrade to ChatGPT’s temporary chat that retains personalization settings while keeping conversations private.
- What Changed: The updated feature now preserves memory, chat history, style, and tone preferences while still preventing conversations from influencing the user’s account.
- Privacy Note: Temporary chats still don’t appear in history or train models, but OpenAI may keep copies for up to 30 days for safety monitoring.
- Age Detection: OpenAI also rolled out an age prediction model that may incorrectly restrict some adult users, who can verify their age to unlock full access.
OpenAI has launched testing last week for an upgrade to ChatGPT’s temporary chat feature that lets users retain personalization settings while keeping conversations private. The updated feature retains personalization including memory, chat history, style, and tone preferences while still preventing those conversations from influencing the user’s account.
In the original temporary chat mode, ChatGPT does not use or reference past conversations or saved memories, even with memory turned on.
How It Works
To use the enhanced feature, users can start a temporary chat by opening a new chat and clicking the pill-shaped ‘Temporary’ button in the top-right corner. The new personalization option can be turned off by users who prefer the original clean-slate approach.
With the update, the conversation itself still remains temporary and does not show up in chat history. However, ChatGPT can respond in a way that matches how users normally interact with the assistant. The opt-out design for personalization retention means the enhanced experience becomes the new default while preserving the clean-slate option for users who want it.
Despite these enhancements, temporary chat still follows Custom Instructions if enabled, but the conversation itself is not remembered. OpenAI’s approach creates a middle ground between completely forgetting user preferences and retaining full conversation context. Users who value both privacy and a personalized experience can now use temporary mode without starting from scratch on their interaction style.
The upgrade addresses a longstanding limitation. Previously, temporary chat forced users to choose between two extremes: either accept ChatGPT’s memory and personalization features would be completely disabled, or skip temporary mode entirely and save the conversation to their history.
The upgrade eliminates this tradeoff by preserving personalization metadata while still discarding the actual conversation content. Users can maintain their preferred communication style, tone adjustments, and accumulated preferences without creating a permanent record of what they discussed.
Privacy Guarantees and Limitations
However, temporary chats do not show up in chat history and are not used to improve OpenAI models. OpenAI may keep a copy for up to 30 days for safety reasons. The retention policy balances user privacy promises with security monitoring requirements, allowing OpenAI to detect and respond to potential abuse while maintaining that conversations are not permanently stored.
While conversations won’t appear in user history or train future models, they’re not immediately deleted. Users seeking absolute conversation ephemerality should be aware of the 30-day retention window. The policy mirrors similar approaches across the AI industry, where platforms balance user privacy expectations with legal and safety obligations.
In practice, privacy-conscious users have increasingly turned to temporary or incognito modes in AI chatbots as a way to ask sensitive questions without building a permanent record. The addition of personalization to temporary mode acknowledges that users want privacy without sacrificing the quality of their AI interactions.
ChatGPT can now remember that a user prefers concise answers or technical depth while still ensuring the actual questions and responses don’t persist in their account. This represents a shift from the feature’s original clean-slate design toward balancing privacy with user experience.
Age Restrictions
Meanwhile, OpenAI rolled out an age prediction model to identify user age based on chat pattern and conversation behavior. The system may incorrectly detect some adult users as teenagers and restrict their accounts. Restricted ChatGPT accounts cannot interact on topics like violence, gore, and viral challenges on TikTok-like platforms.
Fortunately, adult users incorrectly restricted by the age prediction model can verify their age to unlock the full GPT experience. The temporary chat upgrade and age detection rollout both reflect OpenAI’s ongoing effort to add safety controls while expanding personalization features across ChatGPT.
Users caught in false positives will need to provide age verification, adding friction to the experience for those the system misjudges.
