Users discovered around July 19-21, 2026 that conversations they had deliberately shared via Claude’s “share” link feature had been indexed by Google and were publicly discoverable through search, in some cases surfacing sensitive content such as cryptocurrency wallet keys, personal names and addresses, and explicit material. Anthropic said it fixed the underlying issue, though many previously exposed shared-conversation links reportedly remained live and reachable afterward. Similar shared-conversation indexing problems have also previously affected Anthropic itself, as well as OpenAI’s ChatGPT and xAI’s Grok.