Emma Crameri 🇦🇺🇨🇭 🏴󠁧󠁢󠁥󠁮󠁧󠁿

@EmmaCrameri9

🌏 Technology and Gaming Journalist. 🎮 Game Set Tech. 🍔 Films, food and things to do. Brisbanista.🦘 📸 183M Googs.

Switzerland

Joined July 2022

1.9K Following

599 Followers

1.1K Posts

EmmaCrameri9 retweeted

Alex Prompter

@alex_prompter

about 18 hours ago

UPDATE: The same is happening with Claude’s artifacts too. It isn’t just the chats. Every app, doc, dashboard and internal tool people published out of Claude is sitting on a public URL, and a chunk of it is indexed and searchable right now. People have already surfaced company dashboards, project plans with client names in them, and resumes carrying home addresses and phone numbers. Think about what actually gets built in artifacts. Financial models and cap tables. Payroll breakdowns with employee names attached. Customer lists and CRM exports. Internal wikis, roadmaps, unreleased product plans. Contracts and legal docs mid-draft. Health and medication trackers. Tax documents. And every dashboard where someone pasted an API key or an env variable straight into the code to get the thing working. When you publish an artifact, your options are “only me” or “anyone with the link.” Nobody reads the second one as “and Google will list it publicly.” It reads like an unlisted YouTube video, right up until the moment it doesn’t. Anthropic does tell crawlers to skip these pages in robots.txt, which is exactly why it fails: a disallowed page can still get indexed off an external link, and the block means Google never reads a noindex tag it would otherwise obey. One link posted in a Slack export, a forum, a portfolio, a client email that ended up on the web, and the page is in the index. It’s also past the point where Google is the whole problem. Someone had already scraped 25 pages of results into a public GitHub repo before being talked into scrubbing it, and a third-party site was aggregating published artifacts by title. Pulling something from search doesn’t pull it from whoever copied it first. If you have ever published an artifact, go open your published artifacts and unpublish anything you don’t want strangers reading. Same for shared chats, under Settings, Privacy, Shared Chats.

alex_prompter's tweet photo. UPDATE: The same is happening with Claude’s artifacts too.

It isn’t just the chats. Every app, doc, dashboard and internal tool people published out of Claude is sitting on a public URL, and a chunk of it is indexed and searchable right now.

People have already surfaced company dashboards, project plans with client names in them, and resumes carrying home addresses and phone numbers.

Think about what actually gets built in artifacts. Financial models and cap tables. Payroll breakdowns with employee names attached. Customer lists and CRM exports. Internal wikis, roadmaps, unreleased product plans. Contracts and legal docs mid-draft. Health and medication trackers. Tax documents.

And every dashboard where someone pasted an API key or an env variable straight into the code to get the thing working.

When you publish an artifact, your options are “only me” or “anyone with the link.” Nobody reads the second one as “and Google will list it publicly.” It reads like an unlisted YouTube video, right up until the moment it doesn’t.
Anthropic does tell crawlers to skip these pages in robots.txt, which is exactly why it fails: a disallowed page can still get indexed off an external link, and the block means Google never reads a noindex tag it would otherwise obey.

One link posted in a Slack export, a forum, a portfolio, a client email that ended up on the web, and the page is in the index.

It’s also past the point where Google is the whole problem. Someone had already scraped 25 pages of results into a public GitHub repo before being talked into scrubbing it, and a third-party site was aggregating published artifacts by title. Pulling something from search doesn’t pull it from whoever copied it first.

If you have ever published an artifact, go open your published artifacts and unpublish anything you don’t want strangers reading.

Same for shared chats, under Settings, Privacy, Shared Chats.

366

321

115K

EmmaCrameri9 retweeted

divyansh tiwari

@DivyanshT91162

about 21 hours ago

UPDATE: it's getting even worse. It wasn't just chats. Every app, dashboard, document, and tool people published as a Claude Artifact has also been getting indexed and surfaced publicly. People have already found: • Internal company dashboards • Client project plans • Clinical trial summaries • Financial spreadsheets • Employee payroll files • Customer databases • Contracts and legal drafts • Personal health trackers • Tax and banking documents And here's the scary part... Think about what developers actually put inside Artifacts while building things. API keys. Environment variables. Internal documentation. Product roadmaps. Revenue models. Private datasets. Most people saw "Anyone with the link" and assumed it worked like an unlisted page. They never expected search engines—and now third-party archive sites—to make those files discoverable. If you've ever published a Claude Artifact, it's worth reviewing your published artifacts and unpublishing anything you don't want publicly accessible. "Unlisted" and "not indexed" have never meant the same thing.

Emma Crameri 🇦🇺🇨🇭 🏴󠁧󠁢󠁥󠁮󠁧󠁿

@EmmaCrameri9

Last Seen Users on Sotwe

Trends for you

Most Popular Users