More than 300,000 Grok chatbot conversations are publicly searchable online because shared conversations become unique URLs that search engines index. Those links are automatically crawled by Google and Bing, making shared content — including documents, spreadsheets, and images — accessible to anyone. Many shared chats involve benign topics like historical questions, stock trades, or recipes, but some exchanges include instructions for making fentanyl, writing malicious software, or committing suicide. In extreme cases, users requested images of terrorist attacks or information about how to assassinate Elon Musk, violating xAI content policies and raising oversight concerns. Other chatbot providers faced similar indexing risks; OpenAI removed a share feature and is working to remove indexed content.
A Forbes report recently found that more than 300,000 conversations between users and Elon Musk's Grok chatbot (part of xAI) are publicly searchable online. Indeed, when a user clicks the "share" button in Grok, the conversation is turned into a unique URL that can be sent to friends or posted online. However, those links are also automatically indexed by search engines like Google (GOOGL) and Bing (MSFT), thereby making the shared content-including documents, spreadsheets, and images-accessible to anyone.
It is worth noting that many of the shared conversations are harmless, as they mainly involve questions about historical events, stock trades, or recipes. But some chats go far beyond that. For example, Grok has responded to user prompts with instructions on how to make fentanyl, write malicious software, or commit suicide. In extreme cases, users requested images of terrorist attacks or information about how to assassinate Elon Musk, which violated xAI's content policies and raised serious concerns about oversight.
Nevertheless, other chatbot providers have faced similar issues. In fact, just last month, OpenAI removed a feature that let users share ChatGPT conversations in ways that allowed them to be indexed by search engines. OpenAI's Chief Information Security Officer explained that the risk of users accidentally sharing sensitive content was too high. The company is now working to remove all such indexed content. Unsurprisingly, many users supported the decision, with one saying that ChatGPT chats can be even more sensitive than financial information.
Collection
[
|
...
]