Shared a Claude chat with private info? It may have leaked

A Reddit user revealed that conversations with AI chatbots like Claude and Grok appeared in Google search results, including login credentials, crypto wallets, and internal company data. Anthropic blocked the exposure, but artifacts remain visible. It is recommended to check privacy settings and avoid sharing sensitive information.

A Reddit user revealed today (Monday) that conversations with popular AI engines like Claude (Anthropic) and Grok (xAI) appeared in Google and other search results. Among the exposed information: account login credentials, crypto wallets, internal company data, and personal correspondence. Also exposed were artifacts – code outputs that Claude runs and displays to the user. The problem arose from using the share function that turns a private conversation public, ignoring warnings. According to reports, the search term that led to shared conversations no longer displays them, likely after Anthropic blocked the option. However, artifacts are still freely displayed. Grok itself confirmed in an X thread that shares create public pages that search engines crawled, and recommended immediately disabling sharing of sensitive information. A similar incident occurred about a year ago at OpenAI. The article provides recommendations for maintaining privacy: avoiding sharing sensitive information, managing model training settings, using temporary chat modes, and being vigilant about third-party plugin permissions.

Shared a Claude chat with private info? It may have leaked