Elon Musk's Neural Network Allowed User Conversations to Leak onto the Internet

More than 370,000 Grok artificial intelligence chats were posted on the platform and indexed by search engines, making them accessible to all users.

As BAKU.WS reports with reference to Forbes, in addition to text dialogues, user-uploaded photos, tables, and documents were also made publicly available.

It is noted that the culprit was the Share function, which creates a unique link to show the chat to other users. It turned out that these links were indexed by search engines, despite the absence of appropriate warnings.

Forbes journalists discovered that in the published chats, users shared personal data, including names, medical and psychological questions, and in some cases even passwords. Requests for generating illegal content were also identified: instructions for manufacturing prohibited substances, creating malicious software and explosives, as well as materials with calls to violence.

Notably, similar cases were recorded in the summer with other AI platforms, including ChatGPT, however, in those cases, the publication of correspondence occurred with the consent of users.

Grok is a generative AI-based chatbot developed by xAI company. Based on a large language model (LLM), it was launched in 2023 at the initiative of American entrepreneur Elon Musk.