Users of the Reddit and X platforms reported that the Google search engine has started indexing conversations with the Claude chatbot, as well as materials created using this artificial intelligence, including applications, documents, spreadsheets and web pages. According to Axios, confidential correspondence could be detected using a special search query."site:claude.ai/share ". TechCrunch journalists noted that sensitive data was publicly available, including medical records, confidential corporate documentation, as well as personal information of minors, including names and contact phone numbers.

Representatives of the company Anthropic explained that links sent privately are not subject to indexing by search engines. However, when such links are published on public forums or on social networks, they become available to search robots. This technical feature led to an unintended leak of information from users who created links to exchange conversations with others. TechCrunch journalists re-checked the search results at noon on July 27, 2026 and did not find any previously available correspondence, which indicates that the vulnerability was promptly eliminated by the developers.

This incident followed a similar incident that occurred a week earlier, when SEO specialist David Konitzny reported on Google indexing user conversations using the DeepSeek Chinese language model. Most of the correspondence discovered at that time was written in Russian. Such incidents emphasize the importance of caution when using the functions of sharing access to dialogues with artificial intelligence systems and the need for a clear understanding of the indexing mechanisms of published content by search engines to prevent unauthorized dissemination of confidential information.