On July 21, SEO optimization specialist David Konitzny identified a serious security issue in the popular AI platform DeepSeek. It turned out that Google is actively indexing public links to user conversations created through the "Share" function. The search results included not only work documents and financial analytics, but also personal correspondence, including those in Russian.

The leak mechanism is simple: when a user clicks "Share" in DeepSeek, the system generates a unique public link to the conversation. If the owner does not disable indexing in the account settings, these links become available for scanning by search engine bots. As a result, confidential data—from business plans to personal discussions—ends up in the public domain.

The scale of the problem is impressive: tens of thousands of pages have already been found in the search results. Russian-language conversations are particularly alarming, indicating active use of DeepSeek in the Russian segment of the internet. Compromised links can be manually removed through account settings, but this does not guarantee that the data has not already been copied by third-party services.

Situation Analysis

This incident is a classic example of the conflict between convenience and security. The "Share" function is useful for collaboration, but its implementation without indexing control is a gross mistake by the developers. DeepSeek users should immediately check their privacy settings and disable the automatic creation of public links. Otherwise, we may see even larger-scale leaks, especially in the corporate sector.

Expert opinion: DeepSeek is not the first AI service to face such a problem, but the scale of indexing here is unprecedented. This should serve as a wake-up call for all platforms using similar data-sharing mechanisms. Without implementing a "do not index" option by default, such leaks will continue to occur.