Anthropic reported an "increase in errors" for requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 starting at 9:23 AM ET.
The company stated that it identified the cause approximately 15 minutes later and subsequently applied a fix, resolving the issue by 12:16 PM. Additionally, an increase in errors for Claude Sonnet 5 was reported for a short period shortly after noon.
OpenAI reported that as of 10:43 AM, there was an increase in errors and degraded performance for ChatGPT and Codex across. The issue was resolved by 12:55 PM after mitigation measures were implemented.
xAI's Grok displayed an error message to users stating that the model was "experiencing issues." The company stated that it is working to restore service as quickly as possible.
According to reports from DownDetector, the number of reports regarding Grok surged from fewer than 10 just before 9:00 AM to 1,365 by 9:45 AM. As of the time of writing, that number has decreased to 273.
Google has not publicly acknowledged any issues regarding its Gemini models. However, on DownDetector, reports of problems spiked from 23 around 10:30 AM to 412 shortly after 11:00 AM.
Data collected by StatusGator indicates that the Gemini API may have also experienced an outage between 10:45 AM and 11:15 AM. The status subsequently returned to normal.
Source: Four major AI models suffer rare overlapping downtime (Ars Technica AI, 2026-09-04)