Users discovered that private conversations with Anthropic's Claude chatbot could be found via web searches, raising privacy concerns. The issue emerged after a Reddit user flagged that some shared chats, including discussions on political party choices and legal ethics, appeared in search results. This highlights a gap in privacy protections for AI chatbots, despite efforts to restrict access through technical measures.
Anthropic instructed web crawlers not to index shared chats using robots.txt, a standard method for controlling website access. However, search engines like Bing and Google still indexed some pages, as noted by WIRED. Bing showed about 612 results for 'site:claude.ai/share' at the time of the report, while Google no longer displayed these results. Bing advised developers to use both robots.txt and a 'noindex' tag to block indexing, but Anthropic's shared chat pages lacked the 'noindex' tag.
The issue was previously raised in September 2025 when Anthropic faced criticism for the same problem. The company claimed it used robots.txt to prevent crawlers from accessing shared chats, but the effectiveness of this method remains uncertain. Despite this, AI labs like Anthropic, Meta, and OpenAI use robots.txt to block competitors' crawlers from accessing their sites. Google did not comment on the practice of competitors blocking its AI training crawler.
Source: wired
