If you share a Claude chat with someone, you are basically handing them a public web link. That link is just like any other page on the internet - search engines can find it if somebody posts it on a forum or social media.
That is exactly what happened over the weekend. Reddit users noticed that Google had indexed a bunch of these shared Claude links. The conversations and interactive Artifacts were showing up in search results.
This type of exposure is not unprecedented. In 2025, a similar vulnerability with ChatGPT share links allowed a researcher to scrape roughly 100,000 conversations. The core problem is that any publicly posted URL can be discovered by search engines, regardless of whether the platform intended it to be indexed. Anthropic relies on the unguessability of its share links for privacy, but that safeguard disappears once a user posts the link on Reddit, Twitter, or any other public forum.
404 Media was the first news outlet to report it on Monday morning. By Monday afternoon, TechCrunch tested the same search method and found no results, suggesting that the exposure had been remediated.
Get the market news that matters in a five-minute read with Market Briefs, our free daily newsletter
Anthropic said it does not send search engines any directory or sitemap of shared chats. The company spokesperson, Amie Rotherham, said, "These shareable links are not guessable or discoverable unless people choose to share them themselves." But once you share a link publicly - on a forum or social media post, for example - it can get crawled and indexed just like any other public webpage.
Ned Adriance from Google stated that the company provides website operators with clear tools to manage crawling and indexing, and it always respects those directives.
Nobody knows exactly how many conversations were exposed this time. Forbes reported a similar incident last year, during which Google's estimate was that it had indexed fewer than 600 Claude chats before those pages vanished from search. Also last year, 404 Media noted that a researcher managed to scrape around 100,000 ChatGPT conversations from public share links.
The exposed data covered a range of sensitive material. Futurism found a detailed patient medical record, clinical trial data with patient names, a list containing children's names and their phone numbers from a primary school, internal-only company documents, and employee reviews that revealed private worker details. Fortune reported that one exposed chat was labeled "shared by Anthropic" and also showed Claude producing erotica.
The underlying issue has not changed. Anytime you click that share button on a Claude conversation, you are creating a publicly accessible link. If you then post that link anywhere public, it becomes fair game for search engines and third-party archives.
Anthropic recommends checking your own chat privacy settings. You can find them under Settings → Privacy → Shared Chats. That screen shows you which conversations have public links.
Join Market Briefs, our free daily newsletter, for a quick daily rundown of the markets
