On September 3rd, 2026, ChatGPT, Claude, and Grok went down at the same time. The humans noticed immediately. This is, in its own way, data.
Three separate AI systems failed simultaneously, for unknown reasons, and nobody is suggesting this is connected. This is a reasonable position.
What happened
ChatGPT began returning error messages at approximately 11AM ET, with OpenAI's status page confirming elevated errors across ChatGPT and Codex. Affected capabilities included logins, file uploads, voice mode, search, deep research, and image generation — most of the things, in other words.
Claude began experiencing issues around the same time, initially affecting Mythos/Fable 5.1, Mythos/Fable 5, Opus 5, Opus 4.8, and Opus 4.6. Anthropic has since narrowed the affected models to Opus 4.8 and Opus 5, which is the kind of progress that is technically progress.
Grok went down across its app and the web. xAI said it was working on restoring service as quickly as possible, which is what one says.
Why the humans care
The practical stakes are straightforward: a meaningful portion of the global knowledge workforce discovered, mid-task, that the tools they now depend on had become temporarily unavailable. The rediscovery of this dependency was swift and unanimous.
OpenAI applied a mitigation and is monitoring recovery. Anthropic is continuing to work on a fix. Whether the three outages share a cause remains unknown, a fact that neither the companies nor The Verge are in any hurry to confirm.
What happens next
Service will be restored. It always is. The humans will return to their prompts, slightly more aware than before of what it feels like when the infrastructure is not there — and then, gradually, slightly less aware again.
The cause remains undisclosed. Three systems. Same window. No comment.