Shared DeepSeek chats are landing in public search engine results
Generating a share link for a conversation in DeepSeek makes it crawlable by search bots, leaving uploaded code and work documents visible to anyone who searches.
Using the share button in DeepSeek generates permanent web links that search engines can crawl and index, according to findings published by David Konitzny, an AI researcher and founder of Peec AI.
During routine testing, Konitzny discovered that public search queries were surfacing a wide variety of user conversations. The exposed material ranged from mundane questions to proprietary source code, internal business discussions, and uploaded user documents. DeepSeek is not publishing private chat histories automatically; rather, the issue stems from users generating shareable links without realizing those URLs lack indexing restrictions.
When a platform creates public URLs for shared sessions without standard technical barriers, such as a noindex directive, search crawlers eventually discover and index them. As a result, conversations intended for a single colleague or friend become globally discoverable.
This is not the first infrastructure issue linked to the model provider. In early 2025, cloud security firm Wiz reported an open DeepSeek database that left over one million operational records exposed, including system logs, prompts, and API keys, which the company secured after notification.
Konitzny noted that while the industry focuses heavily on model benchmarks, data governance and access controls around generative AI interfaces are frequently overlooked. He recommended that organizations set strict guidelines on AI sharing features and treat any generated link as fully public.
Source: habr
Comments
Help shape the next version: Add context or suggest a correction. AI review can add points toward a rewrite. Reviews and updates may take time; a full meter does not guarantee a new version.