Anthropic компанийн Claude чатботтой хийсэн хувийн ярианууд Google хайлтын системд илэрч, хэн ч үзэх боломжтой болсон нь нууцлалын асуудал дагуулж байна.
Өнгөрсөн амралтын өдрүүдэд зарим хэрэглэгчид Claude-тай хийсэн яриа болон “Artifacts” буюу чатботын үүсгэсэн код, баримт бичгүүд Google-ийн хайлтын үр дүнд гарч ирж байгааг илрүүлжээ. Эдгээр мэдээллийн дунд өвчтөний эмнэлгийн түүх, бизнесийн нууц баримтууд, хувийн утасны дугаар, цахим төлбөр тооцооны мэдээлэл зэрэг эмзэг агуулгууд багтсан байна.
Anthropic компанийн тайлбарласнаар, хэрэглэгчид өөрсдийн чат эсвэл төслийг “хуваалцах” (share) товчийг ашиглан олон нийтэд нээлттэй холбоос үүсгэсэн тохиолдолд ийм нөхцөл байдал үүсдэг байна. Хэдийгээр тус компани холбоосыг хуваалцах үед сануулга өгдөг ч, хэрэглэгчид уг холбоосыг олон нийтийн форум эсвэл сошиал сүлжээнд байршуулснаар хайлтын системүүд тухайн агуулгыг индексжүүлэх боломжтой болдог аж.
Тус компани өөрсдийн нууцлалын бодлогын дагуу чатын лавлах болон сайтын газрын зургийг хайлтын системүүдтэй хуваалцдаггүй гэдгээ мэдэгдлээ. Гэсэн хэдий ч, олон нийтэд нээлттэй болгосон холбоосууд нь бусад вэб контентын нэгэн адил гуравдагч талын үйлчилгээнд архивлагдаж, хайлтаар илрэх эрсдэлтэй хэвээр байна.
Хэрэглэгчид өөрсдийн хуваалцсан чатуудыг удирдахын тулд “Settings” цэсний “Privacy” хэсэгт нэвтэрч, “Shared chats” эсвэл “Shared artifacts” хэсгээс хандалтыг цуцлах боломжтой. Шинжээчид хэрэглэгчдийг чатботтой харилцахдаа хувийн болон эмзэг мэдээллээ хуваалцахгүй байхыг зөвлөж байна.
Дэлгэрэнгүйг эх сурвалжаас харах
Эх сурвалжийг нээх ↓
It turns out Claude may not be the best confidant (well, at least if you’re not careful with your privacy settings). Over the weekend, we got yet another reminder that conversations with chatbots are not guaranteed to stay within the chat box. Some users on Twitter and Reddit showed that conversations with Anthropic’s Claude can be crawled and made searchable on Google, essentially readable by anyone who comes across them.
The conversations that show up in Google search appear to be both conversations and “Artifacts,” which are the lines of code, documents, and even vibe-coded apps that Claude generates in response to a user’s prompting. And because people are not always all that careful about what they put into Claude or other chatbots, these indexed conversations include some pretty personal information.
Futurism reported spotting documents that included a patient’s medical history, clinical trial results, sensitive internal business documents, and a directory of phone numbers for primary school students, among other things. 404 Media found a vibe-coded AI therapy app, meeting notes, and medical billing data. Reports from users on social media claim to have found personal information including home addresses, phone numbers, and some cryptocurrency wallet keys.
So, how do these chats end up available for public viewing via Google search results? Per Anthropic, the issue is the privacy settings that these users have selected. Claude, like other chatbots, has a “share” feature that allows users to pass along their conversations and projects so others can see them. When a user does this, it creates a public link that can be seen by anyone who has access to the URL.
Claude does provide somewhat of a warning for this: when users hit the “share” button, they get a message informing them that they are making that chat public and “anyone with the link can view” it. However, most users probably are not aware of the fact that it could end up getting spotted by a web crawler and made available for literally anyone to see.
Anthropic told Gizmodo that the only way shared links end up in search results is when they have been posted somewhere that search engines can find them, like on a forum or social media. Links sent privately between two people supposedly stay just between those parties—but that’s assuming the other person doesn’t further share the link or post it somewhere.
“We give people control over sharing their Claude conversations publicly, and in keeping with our privacy principles, we do not share chat directories or sitemaps with search engines like Google. These shareable links are not guessable or discoverable unless people choose to share them themselves,” a spokesperson for Anthropic told Gizmodo. “When someone shares a conversation, they are making that content publicly accessible, and like other public web content, it may be archived by third-party services.”
If you have shared a Claude link or if you aren’t sure if you have, you can find it by going to Settings, then Privacy and clicking “Manage” next to “Shared chats” or “Shared artifacts.” From there, you can choose to “unshare” a link and revoke access. At least then no one will be able to click through a Google search result and read your chat, though the indexed part will probably stick around in search for a bit.
This same thing—chats ending up indexed and publicly accessible—has happened in the past with ChatGPT, Claude, and Grok. We’ll probably get another reminder of this in a few months when people once again discover that this is happening. Probably just don’t share your chatbot conversations, and definitely don’t share sensitive information with those chatbots.
