Claude shared chats indexed by search engines, data visible

Shared conversations from Anthropic's Claude AI chatbot were briefly discoverable through public search engines, exposing potentially sensitive user data to a wider audience than intended. The exposure, which became widely known over the weekend of July 25-26, 2026, involved public share links that users had generated to send conversations to others. These links, once posted online or shared publicly, were indexed by search engines, allowing anyone to find them by searching for terms contained within the chat transcripts.

Reports from users on social media platforms indicated that a search query specifically targeting Claude's share domain returned numerous results. The indexed conversations reportedly included a wide range of sensitive material, such as legal discussions, technical work, source code, business information, personal topics, and credentials like API keys and cryptocurrency wallet details. There is no evidence that private, unshared Claude conversations were breached or that the company's systems were compromised. The exposure was limited to conversations for which users had explicitly generated a public share link.

Anthropic's documentation states that...