Experts Advise How to Protect Privacy When Chatting with AI Services

The recent disappearance of Mali and Liel Yahalomi has reignited concerns about privacy when using artificial intelligence services. The two had planned their disappearance carefully, withdrawing cash, deleting apps, and consulting Google’s Gemini AI to learn how to buy an untraceable phone, reduce digital footprints, and identify South American countries without extradition agreements with Israel. Despite deleting their chat histories, authorities managed to recover some conversations and track their whereabouts in Argentina.

This case has raised a widespread question among AI users: how much privacy can one truly maintain when interacting with AI systems without leaving recoverable data? Privacy experts cited by the technology news site ZDNET say it is possible to significantly reduce stored information, but absolute anonymity is difficult to guarantee. Recommendations vary slightly depending on the AI platform.

For ChatGPT, users can interact without logging in or use the Temporary Chat feature, which does not save conversations or use them for model training. Account holders can disable "Improve the model for everyone" in Data Controls and delete existing chats. Users should also be cautious when sharing chat links, as anyone with the link can view the conversation.

Google’s Gemini offers a Temporary Chat mode where conversations are deleted within 72 hours and an option to disable activity saving to prevent chats from being used for training. However, disabling this also removes user access to chat history. Gemini Live users can also turn off voice recording storage, and the Personal Intelligence setting allows disabling the Memory feature that stores personal details for future conversations.

Anthropic’s Claude provides an Incognito mode to prevent chat history saving and privacy settings to disable model improvement contributions and Memory. Shared chat links can be revoked at any time.

Despite these privacy features, deleting chats does not guarantee immediate or complete erasure, as providers may retain data temporarily for security and abuse prevention. The best way to minimize exposure is to avoid sharing unnecessary identifying information, as more generic inputs reduce the risk of personal data leakage.