On the evening of September 3, 2026, local time, OpenAI's ChatGPT and Codex, Anthropic's Claude, and SpaceX AI's Grok experienced large-scale service degradation almost simultaneously. Downdetector data showed over 12,000 fault reports for OpenAI, around 1,200 for Claude, and approximately 1,000 for Grok. The outage lasted for about 3 hours and 40 minutes. By 1:10 AM Beijing Time on September 4, all services had been restored. All three companies confirmed the faults and initiated repairs. Anthropic stated that it had identified the cause of the increased error rate in Claude, while OpenAI said it was conducting investigations. This incident, likely related to issues with shared cloud computing infrastructure, has raised concerns about the stability of AI services.
