AI Pulse · AI Trends Pulse

Over the weekend, Reddit users discovered more than 600 Claude AI share links showing up in Google and Bing search results. These weren’t meant to be public. They contained API keys, resumes, clinical trial data, financial information, and other sensitive material people had fed into Claude conversations or artifacts.

The culprit was a missing noindex tag on Anthropic’s share URLs. When someone clicked “share” on a Claude chat or artifact, the resulting link was technically public but not intended to be crawled. Search engines indexed them anyway. By Sunday, most of the exposed links had vanished from search results, which suggests Anthropic patched the issue. Cached copies still exist, though, and the original report notes that once something is indexed, it takes time to fully disappear.

This is a sharp reminder that “share” does not mean “private”. If you or your team are using any AI tool to draft client proposals, analyse confidential data, or prototype internal systems, assume that anything with a shareable link can leak. It doesn’t matter if the tool is Claude, ChatGPT, or something else. If you generate a link and send it around, you’ve created a vector for exposure.

The fix on your end is simple. Before you click share on an AI chat or artifact, scrub it of anything you wouldn’t post on a public forum. Better yet, set a team rule that client data and proprietary information stay in tools where you control access and indexing. This is exactly the kind of workflow hygiene we build into an AI command centre, where you define what gets shared, where it goes, and who sees it. One missing tag shouldn’t be the difference between a private draft and a Google search result.