Why Was Elon Musk's Grok Chatbot Suddenly Talking Nonsense?

What users saw

Starting midweek, a number of people using Grok — the AI chatbot built by Elon Musk's company xAI — began getting answers that made no sense at all. One user asked for a PDF and got back a jumble of disconnected words, strung together like a broken sentence generator. Another user said they got a wall of links to research papers instead of an actual answer to their question.

How widespread the problem was

It seems to have mainly affected people using the free "Grok Lite" tier via the Grok.com website specifically — not the version of Grok that runs inside X (formerly Twitter), which kept working fine. Journalists who tried to reproduce the bug themselves couldn't reliably trigger it, suggesting it only affected a limited subset of users rather than everyone. Still, it was enough to flood online forums with complaints.

What xAI said

Grok's official account described it as "a rare temporary generation glitch" and suggested that starting a brand-new chat, or having the bot regenerate its last response, usually fixed the problem. Some affected users confirmed this worked; others said they still got the same garbled answers even after refreshing multiple times.

Why this is getting extra attention

This kind of hiccup happens to AI chatbots from time to time and usually isn't a big deal on its own. But it's happening at a time when xAI has reportedly lost a large chunk of its founding team and dozens of researchers and engineers over the past few months — so every visible misstep understandably draws more attention than it otherwise would.