Claude’s sharing feature was built to make it easy for users to send chats, apps, and documents to other people. But over the weekend, that convenience turned into a privacy scare when some of those shared links began showing up in Google search results.

The episode exposed a gap between how users may have understood Claude’s “share” option and how public links can behave once they are visible on the wider web. Reports from Axios and TechCrunch said users were able to surface Claude artifacts — and, at least at one stage, shared conversations — by using Google search operators targeting Claude’s share URLs.¹²

That mattered because the material was not trivial. TechCrunch reported that some shared content allegedly included health records, company documents, and children’s names and phone numbers, underscoring how a tool meant for limited sharing can create broader exposure if people treat a public link as semi-private.² Axios similarly noted that most users probably did not expect material shared with friends or coworkers to become discoverable through a search engine.¹

Anthropic’s position was narrower. The company said it does not provide “chat directories or sitemaps” to Google and argued that shared links become searchable only if users post them somewhere a search engine can crawl, such as a forum or social media site. In the company’s words, “These shareable links are not guessable or discoverable unless people choose to share them themselves,” and “when someone shares a conversation, they are making that content publicly accessible.”¹²

The tension, then, is less about whether the links were technically public than whether users were adequately warned about what “public” could mean in practice. By Monday, both outlets reported signs the exposure had been remediated or at least was no longer easily reproducible in Google results.¹² But the incident adds Claude to a growing list of AI products facing scrutiny over how sharing features collide with ordinary expectations of privacy.