Google has indexed private conversations of DeepSeek users — data leak through the "Share" feature
On July 21, SEO specialist David Konitsny discovered a critical vulnerability in the DeepSeek platform: Google began indexing public links to user conversations. These are pages created through the "Share" function, which became accessible in search results.
Indexed data included work documents, financial analytics, and personal correspondence, including Russian-language conversations. This jeopardizes the privacy of thousands of users who may not have realized their discussions were becoming public.
The leak mechanism is related to the fact that DeepSeek generates unique URLs for each conversation when using the "Share" function. If a user does not disable indexing in their account settings, these links automatically appear in search engines. Google, like other search engines, scans such pages, making them accessible to anyone.
Already indexed links can be removed through DeepSeek account settings, but this does not guarantee complete deletion from Google's cache. For users who actively use the "Share" function, this is a serious signal to reconsider their security habits.
My analysis: This is not just a technical oversight — it is a fundamental flaw in DeepSeek's privacy design. In an era when data confidentiality is becoming a critical asset, the platform must by default block the indexing of sensitive pages. Users, especially corporate ones, should immediately check their settings and reconsider their policy for using the "Share" function. The market is already reacting: such incidents undermine trust in AI platforms, and DeepSeek risks losing ground in the race for data security.