Overview
- On Thursday, Sept. 3, users and monitoring sites reported interruptions to ChatGPT, Anthropic’s Claude and xAI’s Grok in overlapping time windows.
- OpenAI logged a brief incident called “ChatGPT Work Mode High Error Rates” that began at 12:04 UTC and was resolved minutes later, and Anthropic reported elevated error rates affecting Claude Sonnet 5 starting about 12:37 UTC.
- Downdetector and social media showed hundreds to thousands of reports from multiple countries, producing widespread user complaints and disruption to work and study routines.
- Each company posted technical details on its status page, applied mitigations and reported service recovery for most affected functions while continuing to monitor performance.
- No firm evidence links the incidents to a single infrastructure failure or coordinated attack and investigators say root-cause attribution remains open, a reminder of broad public dependence on cloud AI services.