Anthropic AI Chats Exposed on Google via Shared Links
Hundreds of private user conversations and interactive Artifacts created on Anthropic's artificial intelligence chatbot, Claude, became publicly searchable on Google over the weekend. Users discovered the exposure using specific search operators, revealing sensitive health records, company data, and personal information. While search availability was subsequently removed, Anthropic stated that links only appear online if shared by users themselves.
Key points
- Anthropic, the artificial intelligence company behind the chatbot Claude, faced a data exposure after user conversations were indexed by search engines.
- Reddit users discovered the issue over the weekend by typing search operators like 'site:claude.ai/share' into Google, revealing over 200 conversations across at least 25 pages of results.
- Exposed data included sensitive materials such as private health records, corporate documents, and the names and phone numbers of children.
- Anthropic stated that users maintain control over their content, emphasizing that links are unguessable unless intentionally shared by the creators.
- The search indexing was removed over the weekend, though many of the leaked logs had already been archived and shared widely online.
Private user conversations and interactive Artifacts generated on Claude, the artificial intelligence chatbot developed by Anthropic, were found publicly searchable on Google over the weekend. The vulnerability came to light after Reddit users employed specific search operators, such as 'site:claude.ai/share', which surfaced more than 200 conversations spanning at least 25 pages of search results. Some of the exposed chat logs contained highly sensitive data, including private health records, confidential corporate documents, and personal details such as the names and phone numbers of children.
The exposure stemmed from Claude's 'share chat' feature, which generates URLs allowing external viewers to access specific conversations or projects. Although the interface warns that anyone with the link can view the content, it does not explicitly state that such links might be indexed by search engines like Google. By contrast, similar sharing functions on platforms like Google Docs typically prevent public web indexing unless explicitly configured otherwise.
In response to the incident, an Anthropic spokesperson maintained that users retain full control over their data and that conversation links remain unguessable unless actively published by the creators. The company suggested that the exposed links only appeared in search engines because they had been posted previously on public forums or social media platforms. Although the direct search engine availability was resolved over the weekend, numerous chat logs had already been archived and circulated widely across the internet.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.