Follow ZDNET: Add us as a preferred source on Google.
ZDNET's key takeaways
- Claude AI shared chats were surfaced by Google queries, Reddit users reveal.
- Private chats were not leaked, and Anthropic reacted quickly.
- To keep your conversations safe, be cautious about creating share links.
Claude has, once again, become the eye of a cybersecurity and privacy storm, much to the amusement of Reddit users.
It was only back in June that Anthropic, the creator behind the popular Claude AI chatbot, had to revoke access to Fable 5 and Mythos 5 due to a White House directive and concerns surrounding a vulnerability or jailbreak. The vagueness and lack of transparency surrounding this issue, however, are in direct contrast to the latest issue facing Claude's PR team: the exposure of shared chats through a simple Google query.
Also: Claude Opus 5 arrives with near Fable performance at half the price
Claude AI shared chats indexed by Google #
Over the weekend, Reddit users in a Claude-related forum revealed that a simple Google dork (or search query), site:claude.ai/share, could be used to pull up chats from Claude AI, ranging from adult fanfiction to news discussions, the creation of fake book reviews, and conversations about programming.
It appears that Anthropic was made aware of the indexing fallout very quickly, as the Reddit thread indicates the Google dork stopped working soon after, and the lists of Claude AI chats vanished from Google.
However, many of the links that have been shared are still active, and most are decidedly not funny. On the contrary, many of the topics are deeply personal and may have financial or legal repercussions, according to ZDNET's findings. (Some users also claim to have found chats involving cryptocurrency wallets, although we have not been able to confirm this.)
Also: OpenAI's agent breached Hugging Face before an AI defender caught it: What users should do next
The fun might now be over for Reddit forum users, but the situation has highlighted a problem that should not have existed in the first place. This isn't the first time AI chat conversations have been inadvertently leaked online, either, with similar scenarios involving ChatGPT and Grok last year.
Were private chats leaked? #
Thankfully, no.
Claude has a shared chats feature that allows users to share conversations with others, if they want to. By clicking the "Share" button in the upper right corner of your chat, a snapshot of your current discussion is created, and then you can copy the link and share it. This provides access to your conversation, but not your full account or other chats.
It appears, however, that once these links were activated, there was no guardrail in place to stop them from being indexed en masse by Google -- or, potentially, by other search engines.
How do I keep my conversations safe? #
Claude chats are private by default, and so unless you intentionally create a shareable link, your conversations shouldn't be exposed.
To check that none of your conversations are shareable, head to Settings > Privacy > Shared Chats, then tap Manage. If you have previously shared a conversation but would rather it not be exposed to the internet, consider disabling any active sharing links.
However, Anthropic should still address this and ensure that such indexing can't happen again. After all, Claude's sharing function doesn't mean "share and make available to everyone online."
It's still a breach, albeit unintentional, and the company should take ownership of the problem beyond just applying a hotfix. Considering that collections of indexed chat links are now online, this should include disabling them server-side.
*ZDNET has reached out to Anthropic for comment, and we will update when we hear back. *