Status as of 16:23 UTC on 3 September 2026 (9:53 pm IST): Anthropic has marked the Claude incident as resolved, and all Claude services are operational again. OpenAI's ChatGPT and Codex are recovering after a fix was applied. xAI's Grok status page still shows an open outage with no update since 13:30 UTC. This post is being updated.
OpenAI's ChatGPT, Anthropic's Claude and xAI's Grok all broke within roughly 90 minutes of each other on Thursday, 3 September 2026. All three companies confirmed incidents on their own dashboards, which is what makes this different from the usual Downdetector noise — three rival labs, three separate stacks, one window.
Which Services Went Down
Anthropic listed claude.ai, the Claude API, Claude Code and Claude Cowork as partial outages on the Claude status page. It's 13:50 UTC update named seven affected models: Mythos 5.1, Fable 5.1, Mythos 5, Fable 5, Opus 5, Opus 4.8 and Opus 4.6. Live user reports for the platform are tracked on our Claude status page at Downbits.
OpenAI's status page flagged 15 affected ChatGPT components and four Codex components under an incident opened at 14:58 UTC. That is more than 90 minutes after Anthropic and xAI had already gone public, so ChatGPT users spent the first half of this outage with no official confirmation at all — the ChatGPT report feed on Downbits was ahead of the dashboard.
xAI logged a model's outage for Grok on the web at 13:30 UTC, saying only that Grok "is experiencing issues". That remains the only thing xAI has said. Reports are still coming in on the Grok status page.
Google's Gemini drew a report spike, too, but Google never confirmed an outage, and Gemini stayed usable throughout.
Outage Timeline
| UTC | IST | Event |
|---|---|---|
| 12:37 | 6:07 pm | Anthropic opens a separate Claude Sonnet 5 error incident (resolved 12:56 UTC) |
| 13:26 | 6:56 pm | Anthropic starts investigating errors on Mythos 5.1, Fable 5.1 and Opus 5 |
| 13:30 | 7:00 pm | xAI posts a model's outage for Grok on the web |
| 13:50 | 7:20 pm | Anthropic publishes the full list of seven affected models |
| 14:58 | 8:28 pm | OpenAI opens its ChatGPT and Codex incident |
| 15:25 | 8:55 pm | Anthropic says only Opus 4.8 and Opus 5 remain affected |
| 16:06 | 9:36 pm | Anthropic deploys a fix and monitors recovery |
| 16:16 | 9:46 pm | Claude's impact ends |
| 16:23 | 9:53 pm | Anthropic marks the incident resolved; Grok's outage is still open |
Is Microsoft Azure the Cause?
Nobody has confirmed it. None of the three companies has named a root cause, and the Azure explanation now spreading across coverage rests on a single circumstantial fact: Azure saw its own Downdetector spike, and all three platforms use Azure to some degree, as 9to5Google noted.
Microsoft has published no Azure incidents for this window. Third-party monitor StatusGator recorded Azure as operational at 15:13 UTC, in the middle of the outage. A Downdetector spike counts user complaints, not confirmed faults — and complaints about a cloud provider always climb when anything built on it breaks.
There is a second detail that cuts against a single shared trigger. Anthropic opened and resolved a completely separate Claude Sonnet 5 incident between 12:37 and 12:56 UTC, before this one started, and Sonnet 5 was not on the list of seven models affected by the main incident. That looks like two unrelated faults at one vendor on one afternoon. We'd treat the Azure theory as unproven until Microsoft or one of the three labs says otherwise. A shared failure window is suggestive; it is not evidence.
How Many People Were Actually Affected
Downdetector logged more than 35,000 US reports for ChatGPT, over 1,400 for Claude, and more than 1,200 for Grok, according to India Today's tally. Those numbers skew heavily American, and they undercount India badly — Indian users hit this outage between roughly 7:00 pm and 10:00 pm IST, right in the middle of the working evening.
If one of these tools failed for you today, tell us which and where you were:
- Report ChatGPT status — is it loading for you now?
- Report Claude status — Anthropic says resolved; we want to know if that matches reality on your connection
- Report Grok status — xAI still hasn't updated, so user reports are the only live signal
Reports take one tap and no sign-up. Say it's working if it's working — a tracker is only worth reading if the "up" votes are honest too.
Quick Answers
Is ChatGPT down right now?
ChatGPT is recovering as of 16:23 UTC on 3 September 2026. OpenAI applied a fix, and error rates are falling, though the incident had not been marked as resolved at that time.
Is Claude back up?
Yes. Anthropic marked the Claude incident resolved at 16:23 UTC on 3 September 2026 and says the impact ended at 16:16 UTC. All Claude services show as operational.
Why did ChatGPT, Claude and Grok go down together?
No cause has been confirmed. OpenAI, Anthropic and xAI have each acknowledged errors without naming a shared trigger, and Microsoft has not reported an Azure incident for the same window.
What To Do While Services Recover
- Switch models rather than apps — Claude Sonnet 5 recovered before Opus 5 did, and Codex was affected separately from ChatGPT.
- Retry failed API calls with backoff instead of re-running whole jobs.
- Watch the vendor status pages for the resolution mark, and user reports for what's actually happening in between.
- Gemini, DeepSeek and Perplexity stayed up through this window if you need a fallback.