An unusual glitch on xAI‘s Grok.com surfaced this week, causing the Grok chatbot to respond with gibberish to some users. After asking the model to generate a PDF, one user received a reply like “match it without and your they and two for planets can practical and often cheese…” with nonsense continuing for several paragraphs. Another user reported that source links in the output pointed to reinforcement learning research sites. Affected users who spoke to TechCrunch said they were using Grok Lite and noticed the problem as early as Wednesday morning. TechCrunch could not reproduce the issue in its own testing, suggesting it likely affects only a small subset of users. xAI did not respond to a request for comment.

Refresh the session copied normal behavior for the closest, but some users said gibberish continued even after multiple refreshes. The glitch appeared to be limited to direct queries on Grok.com; the Grok account on X.com was unaffected. Responding to unhappy users on X, the Grok account confirmed the issue, writing: “That pure word salad is a rare temporary generation glitch. Official status at https://status.x.ai shows all Grok services fully operational with no incidents. Start a fresh chat or regenerate—it usually clears right away. Sorry about the gibberish.” That message came Thursday morning.

The incident follows a period of significant staff turnover at xAI: The Information reported in May that the company had lost most of its founding team and at least 50 researchers and engineers. xAI released its most recent foundation model in July, describing it as “an Opus-class model, but faster, more token-efficient and lower cost.”