The artificial intelligence company of Elon Musk, XAI, has published the transcripts of hundreds of thousands of conversations between its Chatbot Grok and the Bot users, in many cases, without the knowledge or permission of these users.
Every time a Grok user clicks on the “Share” button on one of his chats with the bot, a unique URL is created that allows him to share the conversation by email, text message or other media. However, without the users knowing it, that unique URL is also available for search engines such as Google, Bing and Duckckgo, allowing anyone on the web to look for it. In other words, in Musk Grok, by clicking on the “Share” button, the conversation will be published on the Grok website, without prior notice or discharge of responsibility for the user.
Today, a Grook chats search on Google shows that the search engine has indexed what Google estimates in more than 370,000 user conversations with the Bot. The shared pages revealed conversations between Grok and LLM users, which cover from simple business tasks such as writing tweets to generating images of a fictitious terrorist attack in Kashmir and try to hack a cryptocurrency wallet. Forbes He analyzed conversations in which users asked intimate questions about medicine and psychology; Some even revealed the name, personal data and at least one password shared with the Bot by a Grok user. You could also access image files, spreadsheets and some text documents uploaded by users through Grok’s shared page.
Among the indexed conversations were some initiated by British journalist Andrew Clifford, who used Grok to summarize the newspapers of the newspapers and write tweets for his Sentinel Current website. Clifford declared a Forbes He did not know that by clicking on the button to share, his message would be in Google. “I would bother me a little, but there was nothing that should not be,” said Clifford, who now uses Google’s gemini artificial intelligence.
However, not all conversations were as benign as Clifford’s. Some were explicit, intolerant and infringed Xai’s norms. The company prohibits the use of its bot to “promote critical damage to human life” or “develop biological, chemical or mass destruction.” However, in published and shared conversations, easy to find through a Google search, Grok offered users instructions on how to manufacture illegal drugs such as fentanyl and methamphetamine, codify self -demand malware, build a pump and suicide methods. Grok also offered a detailed plan for the murder of Elon Musk. Through the “Share” function, the illegal instructions were published on the Grok website and were indexed by Google.
XAI did not respond to a detailed request for comments.
Xai is not the only startup that has published the conversations of its users with their chatbots. Earlier this month, Openai’s Chatgpt users were alarmed when discovering that their conversations appeared in Google’s search results, despite having opted for others to see them. However, after protests, the company quickly changed its policy. Openai’s Information Security Director Dane Stuckey described indexation as an ephemeral experiment and declared in an X post that would be discontinued because “it offered too many opportunities for people to accidentally share information that did not intend.”
Read more:
XAI of Elon Musk published hundreds of thousands of chatbot grok conversations
After Openai canceled his function to share, Musk turned around the victory. Grok’s X account then said that he did not have that function, and Musk tweeted in response: “Grok, finally!” It is not clear when Grok added the function to share, but X users have been warning since January that Google was indexing Grok’s conversations.
Some of the conversations in which Grok was requested of instructions on how to manufacture drugs and bombs were probably initiated by security engineers, Redteamers or trust and security professionals. But in at least some cases, Grok’s shared use configuration even confused professional AI researchers.
Nathan Lambert, a computational scientist at the Allen Institute, used Grok to create a summary of his blog’s entries and share it with his team. He was surprised to know, through Forbes, that his message in Grok and the AI response were indexed on Google. “I was surprised that Grok’s chats shared with my team will automatically index on Google, despite not having received any warning, especially after the recent conflict with ChatgPT,” said the investigator resident in Seattle.
Google allows website owners to choose when and how to index their searches. “The editors of these pages have total control over their indexation,” said Ned Adriance, Google spokesman, in a statement. Google already allowed to index the chats with their AI chatbot, Bard, but eliminated them from searches in 2023. Meta continues to allow search engines to detect their shared searches, Business Insider reported.
The opportunists are beginning to notice and take advantage of Grok’s published chats. On LinkedIn and the Blackhatworld forum, marketing professionals have talked about creating and sharing conversations with Grok to increase visibility and recognition of their companies and products in Google search results. (It is not clear how effective these initiatives would be). Sable Kumar, executive director of the Seo Pyrite Technologies agency, showed Forbes How a company used Grok to manipulate the results of a search for companies that would write its doctoral thesis.
“All chats shared in Grok are totally undexable and searchable on Google,” he said. “People are using active tactics to position these pages in the Google index.”
This article was originally published by Forbes Us.
You may be interested: why the 100 -Mdd cannabis empire of Cheech and Chong is no joke
