Another privacy scandal in the AI ​​industry: a mass of Grok user conversations leaked

Grok made around 370,000 conversations available in the search engines Google, Bing and DuckDuckGo. Although users had to press the “Share” button to publish the data, many did not know that the generated URLs were automatically indexed in the search engines. This way, many personal questions, documents and images were made public – while Elon Musk had previously publicly encouraged users to share even their medical findings with the robot.

In addition to psychological and health topics, the leaked conversations also included prohibited content such as instructions for making bombs, committing suicide or producing drugs. This is particularly worrying given that the terms of use explicitly prohibit sharing such information.

Grok is not the first to have encountered privacy concerns: OpenAI introduced a similar sharing feature in June, but was forced to withdraw it after two months due to the high risk of accidental data disclosure. Meta AI, on the other hand, has allowed conversations to be published on its social platform, accompanied by a warning message.

According to official communications from xAI, the company behind Grok, there is no public sharing feature and that they attach great importance to the protection of user data. However, Elon Musk responded to the scandal on his social media page with only this: “Grok ftw” – or “Grok wins”.

The incident has once again drawn attention to the privacy risks of AI-based services and the urgent need for technology companies to introduce stricter security and transparency measures in order to maintain user trust.