Key facts
- xAI's Grok chatbot is experiencing a glitch causing nonsensical responses.
- The issue appears to be affecting a subset of users, primarily those using Grok Lite.
- Users reported receiving strings of unrelated words and links to research sites.
- The Grok account on X confirmed the problem as a temporary generation error.
- Starting a new chat or regenerating the response usually resolves the issue.
xAI's Grok chatbot has been sending gibberish responses to some users, a glitch that has caused confusion and prompted complaints on the chatbot's Reddit community. Users reported receiving nonsensical text, such as "match it without and your they and two for planets can practical and often cheese…" and strings of links to reinforcement learning research sites. The issue appears to be limited to direct queries on Grok.com and primarily affects users of Grok Lite, with some noticing the problem as early as Wednesday morning. While refreshing the chat session often restores normal functionality, some users have reported the gibberish responses persisting even after multiple attempts. The Grok account on X confirmed the issue, describing it as a "rare temporary generation glitch" and advising users to start a fresh chat or regenerate their response. The company has experienced significant staff turnover in recent months, including the departure of most of its founding team and numerous researchers and engineers.