started · updated
Anthropic's Claude AI share links indexed by Google, exposing user data
In July 2026, hundreds of conversation links created with Anthropic’s Claude AI chatbot appeared in Google and other search‑engine results. The links, generated through Claude’s “share” feature, are public URLs that anyone with the link can view. Because the pages lacked a no‑index tag, search crawlers indexed them despite the site’s robots.txt, allowing the URLs to be discovered via queries such as site:claude.ai/share. Reddit users were the first to spot the issue, reporting that the results displayed personal data, medical records, corporate documents, legal advice and even cryptocurrency‑wallet seed phrases. Anthropic said the exposure was not a security breach; the links become public only when users choose to share them and the company does not provide chat directories or sitemaps to search engines. The company subsequently added the missing no‑index tag and removed the pages from Google’s index, although URLs that users had already saved remain accessible. The incident has raised renewed scrutiny of AI‑chat privacy controls.
Entities
Anthropic · Anthropic PBC · Bing · Claude · Claude (AI assistant) · Claude (AI chatbot) · Claude AI · Claude Max · Google · Google LLC · Microsoft Bing · Microsoft Entra ID
Claims
What the coverage asserts, and how many sources carry each claim.
- [○ 1 SOURCE] Anthropic released Claude Opus 5 on 24 July 2026 across its platforms, Amazon Bedrock, Google Cloud and Microsoft Foundry.
- [○ 1 SOURCE] Anthropic published a tutorial video showing how to configure SSO for Claude using Microsoft Entra ID.
- [● 4 SOURCES] Exposed chats included legal advice, engineering code, personal health records, and cryptocurrency‑wallet seed phrases. www.all-about-security.de · cybersecuritynews.com · gizmodo.com · hackread.com
- [○ 1 SOURCE] Opus 5 achieved a 43.3 % score on the Frontier‑Bench agent‑coding benchmark, outperforming Fable 5 and GPT‑5.6.
- [○ 1 SOURCE] Requiring SSO for Claude accounts reduces shadow AI risks by enforcing managed Microsoft credentials.
- [● 9 SOURCES] The indexed pages lacked a proper no‑index tag, allowing search engines to index them despite robots.txt restrictions. www.all-about-security.de · cybersecuritynews.com · gizmodo.com · www.wired.com · www.batista70phone.com · +4 more
- [● 9 SOURCES] Hundreds of Claude AI shared chat URLs were indexed by Google and appeared in public search results. www.all-about-security.de · cybersecuritynews.com · gizmodo.com · www.wired.com · www.batista70phone.com · +4 more
- [● 9 SOURCES] Google results for the affected Claude share pages largely disappeared by the following Sunday. cybersecuritynews.com · www.all-about-security.de · gizmodo.com · www.wired.com · www.batista70phone.com · +4 more
- [○ 1 SOURCE] Reddit users reported the indexing issue on 27 July 2026. www.all-about-security.de
- [● 9 SOURCES] Anthropic had not issued a public statement about the incident at the time of reporting. cybersecuritynews.com · gizmodo.com · www.wired.com · www.batista70phone.com · www.all-about-security.de · +4 more
- [● 9 SOURCES] Reddit users first reported the indexing issue on 25‑27 July 2026. gizmodo.com · www.wired.com · www.batista70phone.com · www.all-about-security.de · hipertextual.com · +4 more
- [● 9 SOURCES] Exposed chats included legal advice, source code, personal health records, business documents and cryptocurrency wallet seed phrases. gizmodo.com · www.wired.com · www.batista70phone.com · www.all-about-security.de · hipertextual.com · +4 more