How to Make Sure Your Claude Chats Don't Get Shared Publicly (Again)

If you usage Claude, Anthropic's AI chatbot, you mightiness beryllium alert that your chats whitethorn person been exposed connected Google. The news collapsed complete the weekend, arsenic Redditors discovered a wealth trove of conversations betwixt Claude users and their AI bots. All 1 had to do was hunt for a keyword followed by "site:claude.ai/share" to find Claude conversations astir that topic. It was, to beryllium frank, rather the mess.

Luckily, Anthropic quickly plugged the information hole. Ever since, searching Google pinch "site:claude.ai/share" returns nary results, truthful you can't find immoderate of these chats coming if you try. However, it's not needfully a immense comfortableness to cognize that Google indexed these conversations successful the first place—and for the 2nd clip successful a year, astatine that. Anthropic dealt pinch a akin business backmost successful September, and while it isn't alone—Google erstwhile besides indexed ChatGPT conversations—you mightiness understandably beryllium concerned astir early chats leaking onto the web, arsenic good arsenic wonderment whether your existent chats were exposed.

Why were Claude chats exposed?

The point is, it doesn't look for illustration each Claude chat had an adjacent consequence of vulnerability here. Anthropic is adamant it doesn't stock chat links to Google directly; instead, the immense mostly (if not the entirety) of these chats look to beryllium conversations that individuals decided to stock pinch others connected the unfastened web, according to Google. Claude lets you stock your chats publically if you wish, and while that tin beryllium convenient for showing friends, family, aliases colleagues a circumstantial discussion, it mightiness besides unfastened up your chat to Google's web crawlers, arsenic erstwhile you stock a chat, it generates a nationalist link.

You May Also Like

So, there's a metallic lining successful that backstage Claude chats that weren't shared are improbable to beryllium progressive successful this latest exposure. If you don't stock your chats pinch anyone, you should beryllium successful the clear—though Futurism did report that 1 personification successful September's Claude debacle swears they ne'er shared their exposed chats. However, if you do stock Claude chats, your conversations mightiness person been portion of the indexing that happened complete the weekend.

Further complicating things, it seems that Claude's "artifacts" are still visible connected Google. Claude generates artifacts erstwhile the contented of your task meets definite criteria, and tin see thing for illustration documents, codification snippets, HTML sites, SVG images, diagrams, flowcharts, and Interactive React components. In galore cases, users whitethorn want these artifacts to beryllium public, arsenic they whitethorn beryllium meant to beryllium shared widely. However, if you don't want your artifacts indexed connected Google, you'll person to return action.

How to protect your Claude chats from early exposure

While location doesn't look to beryllium immoderate reliable measurement to place whether your chats and artifacts were included successful this exposure, you tin guarantee that nary of them are ever indexed by a hunt motor successful the future. To do so, you'll fundamentally person to unopen down sharing.

To start, caput to Settings > Privacy successful Claude, past take "Manage" adjacent to "Shared chats." Once this caller page appears, you'll spot each shared chats, including a nexus to the speech and the day shared. From here, scroll done the database and prime "Unshare" adjacent to each chat you'd for illustration to remove. This won't delete your chats; only their snapshots connected to these nationalist links.