Anthropic has pushed back on suggestions it was responsible for the exposure of shared Claude chats and Artifacts on Google, telling TechCrunch that share links only surface in search results if users themselves post them somewhere search engines can crawl, such as a forum or social media post, and that privately sent links remain hidden. A company spokeswoman said Anthropic does not submit chat directories or sitemaps to search engines and that the links are not guessable, meaning any public exposure stemmed from users choosing to share content publicly rather than a flaw in the system. Google likewise said it does not control which pages site owners make public, noting the affected pages were indexed across multiple search engines, not just its own.

The exposure, first spotted by a Reddit user on Saturday and reported by 404 Media on Monday, saw shared Claude links containing sensitive material—including a real patient's medical report, clinical trial data with patient names, children's contact details, internal company documents and staff reviews—turn up in Google searches using operators like "site:claude.ai/share". Fortune additionally reported finding a chat labelled "shared by Anthropic" in which Claude had generated erotica, despite this being explicitly barred under Anthropic's usage policy; how that content was produced remains unclear, and Anthropic has not yet commented on the specific case. By Monday afternoon, TechCrunch's repeat searches turned up no results, suggesting the exposure had since been remediated.