Claude AI chat leak exposes privacy risks — lock down your AI conversations

This week, news broke that conversations from Anthropic's Claude AI could be accessed through a Google dork — an advanced search query — because links to shared chats were being indexed by search engines. Anthropic quickly patched the exploit to disable indexing, but the incident raises a broader question: how private are your AI conversations? The answer, according to ZDNET, depends on the user. AI companies do offer privacy controls, though they're often buried in settings.

ZDNET's article outlines specific steps for four major chatbots. For ChatGPT, you can chat without an account entirely, use Temporary Chat (which prevents saving history), turn off "Improve the model for everyone" under Settings > Data controls to stop training on your data, and be cautious when sharing conversations via social platforms. The advice is pragmatic: even with these safeguards, avoid sharing anything too sensitive. The core takeaway: privacy defaults aren't automatic — you need to actively enable them across OpenAI, Anthropic, and other platforms to keep your AI conversations from becoming public.