An unusual glitch has resulted in xAI's Grok chatbot sending nonsensical responses to many users. One user who asked the model to generate a PDF received the response: "match it without and your they and two for planets can practical and often cheese…" with similar gibberish continuing for several paragraphs.
Affected users who spoke to TechCrunch said they were using Grok Lite and noticed the issues from Wednesday morning. One user who checked source links found a series of links to research sites on reinforcement learning. TechCrunch was unable to reproduce the issue in its own tests, and it is likely affecting only a small group of users.
The result has been a considerable number of confused Grok users. The chatbot's Reddit community has been inundated with complaints. Refreshing the session often restored normal functioning, although some users reported that nonsensical responses continued even after multiple refreshes.
The glitch appears to be limited to direct queries on Grok.com. The Grok account on X.com has not been affected. In response to disgruntled users, the Grok account confirmed the issues on X: "That pure word salad is a rare and temporary generation glitch. Official status at https://status.x.ai shows all Grok services running fully without incidents. Start a new chat or regenerate—generally resolves right away. Apologies for the nonsense," the account wrote Thursday morning. xAI did not respond to a request for comment.
xAI has experienced significant staff turnover in recent months, losing most of its founding team and at least 50 researchers and engineers, according to a May report by The Information. The company launched its latest base model in July, describing it as "an Opus-class model, but faster, more token-efficient and lower-cost".
