More AI chat tools are promising stronger privacy for user chats

In short: More AI chatbot services are adding privacy features, but many still rely on policy promises or paid plans, not hard technical limits.

What's going on

Many people now use AI chatbots like ChatGPT, Claude, and Gemini for personal topics, including health, money, and relationships. Privacy experts warn that these chats can be saved and later accessed by the company, partners, contractors, or by law enforcement through legal requests.

WIRED reports that some stronger protections exist, but they are often aimed at businesses. One example is “zero data retention” (ZDR), which is a contract setting that says chat records should be deleted right after the system answers. OpenAI, Anthropic, and Google offer ZDR options for enterprise accounts, but WIRED notes there can still be exceptions, such as limited logging to detect abuse.

For everyday users, many “private” chatbots mainly offer promises not to store chats. WIRED points to services like Proton’s Lumo, Venice.ai, and DuckDuckGo’s Duck.ai. Some of these tools act like a middleman, sending your question to a larger AI service while trying to hide identifying details, but experts say your words can still reveal who you are (like giving your neighborhood in a question).

A smaller group of tools try to add technical barriers. Signal creator Moxie Marlinspike’s Confer uses a “trusted execution environment” or TEE, which is like a locked room inside a computer chip where the chat is processed. WIRED also describes similar approaches from Meta in WhatsApp and from Apple with Private Cloud Compute.

What to watch

Privacy protections may come with trade-offs, including higher prices and weaker performance for tools that run entirely on your own device. Experts also say users should read privacy claims carefully and assume sensitive chats could be stored unless a tool clearly limits that in a verifiable way.

Source: Wired