Enterprise DNA
News AI News

Claude's "share conversation" links are getting indexed by search engines, exposing private chats.

Users found that share URLs, despite Anthropic's robots.txt disallow, end up indexed once posted anywhere external, exposing API keys, crypto wallet.

Enterprise DNA |
Claude's "share conversation" links are getting indexed by search engines, exposing private chats.

AI Pulse · AI Trends Pulse

The play

Never post a Claude share link publicly, even expired ones get indexed and expose API keys, wallet addresses, and business details in search results.

Claude’s shareable conversation links are showing up in Google search results, and they’re exposing things users never meant to make public. API keys, cryptocurrency wallet details, resumes, internal business conversations. All sitting there, indexed and searchable.

Here’s what happened. When you share a Claude chat, Anthropic generates a public URL. The company’s robots.txt file tells search engines not to crawl those links. But that only works if the link stays on Claude’s domain. Once someone posts that URL to Reddit, a blog, a forum, anywhere external, Google treats it like any other link and indexes it. The original report notes this isn’t new. A similar incident happened in September 2025, so this is a pattern, not a one-time slip.

Anthropic says the exposure came from users sharing links publicly, not from crawlers ignoring the robots.txt rules. Technically true, but it misses the point. Most people don’t understand that “share” means “make permanently public and searchable.” They think it’s more like a private Dropbox link. It’s not.

What to do about it

If you or your team use Claude, audit what’s been shared. Search your company domain plus “claude.ai/share” in Google. If anything sensitive turns up, contact Anthropic to request removal and submit a Google de-indexing request. Going forward, treat any share link as fully public. If it contains anything you wouldn’t post on your homepage, don’t share it.

This is exactly the kind of risk we build guardrails for in the Omni Command Centre. You want a layer that flags sensitive data before it leaves your environment, whether through a share link, an export, or a careless paste. The tools are powerful, but they don’t know what’s confidential to your business. You have to tell them, or build systems that do.

More like this?

Free daily email

Get this every morning.

This brief is one item from today's AI Pulse, the short daily read we run for ourselves on what is actually happening in AI. Subscribe free and it lands in your inbox each morning.

Free daily email

Subscribe to the daily AI Pulse

One short read every morning on what is actually happening in AI. Free.

One email a day. Unsubscribe any time.

Working With Claude field guide cover

Free Resource

Put what you just read to work

The free 32-page Working With Claude guide: the full ecosystem, Claude Code, and how to roll it out across a business.

Add your name (optional)

No spam. Unsubscribe any time.