Grok Chatbot Experiences Widespread Glitch, Generating Nonsensical Responses For Some Users
aiinsider ·
xAI's Grok chatbot recently started producing incoherent, gibberish responses for a subset of users, with some receiving unrelated links to reinforcement learning research instead of normal answers. The issue mainly affected those using Grok Lite, and TechCrunch could not reproduce the problem in its own testing, suggesting limited impact. Users on Reddit reported mixed results, with refreshing sometimes fixing the issue while other cases persisted. The glitch was confined to direct queries on Grok.com, leaving the Grok account on X unaffected. The official Grok account on X acknowledged the problem as a rare temporary generation issue, noted that its status page showed everything operating normally, and advised users to start fresh chats. This incident follows significant staff turnover at xAI, including the loss of most of its founding team and at least 50 researchers and engineers since earlier this year. The company had released its latest model, Grok 4.5, in July, describing it as comparable to an Opus-class model but faster and more cost-efficient.
aiinsider