TLDR
ChatGPT, Claude and Grok failed within the same hour on Thursday US time, just after midnight in Sydney. OpenAI's status page listed elevated errors across ChatGPT and Codex and Downdetector passed 35,000 reports. Anthropic's fault on its top models ran from 13:26 to 16:23 UTC and it called the cause an infrastructure issue. xAI confirmed a Grok fault. Gemini and Perplexity stayed up. No company has named a cause, and Azure's status page listed no active events.
KEY TAKEAWAYS
ChatGPT, Claude and Grok all failed within the same hour on Thursday morning US time, just after midnight on Friday in Sydney. None of the three companies has said what caused it.
OpenAI's status page listed "elevated errors across ChatGPT and Codex" from about 10:30am New York time.[1][2] "User reports indicate problems with OpenAI since 10:56 AM EDT," Downdetector posted on X at 12:56am Sydney time. Its count passed 35,000 reports at the peak, about 80 per cent of them about ChatGPT and the rest spread across the app, the API and logins.[2] "We have applied the mitigation and are monitoring the recovery," OpenAI's page said, and spokesperson Kathleen Chaykowski repeated the line to Newsweek.[2] "We are still experiencing degraded performance," the same page read when Gizmodo checked it an hour later.[4]
What each company has confirmed
"We are investigating elevated errors on requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5," Anthropic's status page said at 13:26 UTC, and 24 minutes later it added: "An exhaustive list of affected models: Mythos/Fable 5.1, Mythos/Fable 5, Opus 5, Opus 4.8, Opus 4.6." The incident was marked resolved at 16:23 UTC: "Impact has ended as of 9:16 PT / 16:16 UTC."[3][9] "Claude is mostly back up after an infrastructure issue caused a partial outage across Claude.ai, Claude Code, Claude Cowork, and the Claude API," a spokesperson told Gizmodo. "We know Claude is part of how many of you get work done, and we appreciate your understanding."[4]
xAI confirmed a Grok service issue on its dashboard and was still showing it after OpenAI and Anthropic had moved to recovery.[5] "User reports indicate problems with Google Gemini since 11:15 AM EDT," Downdetector posted on X, and Google confirmed no outage.[2][5] Perplexity users reported no problems.[4]
The night, hour by hour
None of the three companies has published a post-incident report. The record is their status pages, Downdetector's automated alerts and the timestamps on each. In Sydney time, Anthropic's fault began an hour before OpenAI's and was closed twenty minutes before it.
| Sydney | UTC | Who | What they said |
|---|---|---|---|
| 10:37pm | 12:37 | Anthropic | Investigating elevated errors on Claude Sonnet 5. Resolved 19 minutes later. |
| 11:26pm | 13:26 | Anthropic | "We are investigating elevated errors on requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5." |
| 11:50pm | 13:50 | Anthropic | "An exhaustive list of affected models: Mythos/Fable 5.1, Mythos/Fable 5, Opus 5, Opus 4.8, Opus 4.6." |
| 12:30am | 14:30 | OpenAI | Users begin reporting blank responses and stalled prompts across ChatGPT, per Newsweek. OpenAI's page later lists "elevated errors across ChatGPT and Codex". |
| 12:56am | 14:56 | Downdetector | "User reports indicate problems with OpenAI since 10:56 AM EDT." Reports pass 35,000 at the peak, about 80 per cent of them about ChatGPT. |
| 1:12am | 15:12 | Downdetector | "User reports indicate problems with Cloudflare since 11:12 AM EDT." |
| 1:15am | 15:15 | Downdetector | "User reports indicate problems with Google Gemini since 11:15 AM EDT." Google confirms no outage. |
| 1:25am | 15:25 | Anthropic | "The only affected models right now are Opus 4.8 and Opus 5. The rest of the models have recovered to baseline error rate." |
| 1:49am | 15:49 | OpenAI | "We have applied the mitigation and are monitoring the recovery." xAI's dashboard still shows a Grok service issue. |
| 2:06am | 16:06 | Anthropic | "A fix has been deployed for Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 and we are monitoring for recovery." |
| 2:23am | 16:23 | Anthropic | "The issue affecting Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 has been resolved. Impact has ended as of 9:16 PT / 16:16 UTC." |
| 2:55am | 16:55 | OpenAI | "The issue causing elevated errors across ChatGPT and Codex has been resolved. Some Codex remote control users may need to pair their mobile device again." |
Sources: status.claude.com incident log, status.openai.com history, Downdetector's X account, Newsweek. Times converted from UTC to AEST.
What is known about the cause
No company has named a cause. Anthropic's word for it was an "infrastructure issue". "User reports indicate problems with Cloudflare since 11:12 AM EDT," Downdetector posted 16 minutes after its OpenAI alert. Microsoft Azure drew its own spike in reports in the same hour, and OpenAI, Anthropic and xAI all run on Azure among their cloud providers.[5] Azure's status page listed no active events on Thursday: "There are currently no active events."[6] Amazon Web Services also drew user reports in the same window.[8] Cloudflare's status page shows only short regional incidents during the day, none of them in the United States.[7] Cursor confirmed an outage of its own in the same period.[5]
What people saw
These were among the most shared posts on X in the first hour, with their reach at the time Bushletter captured them.
It's not just Grok. EVERY major AI foundational model is down right now. Skynet happening in real-time?
Mark Kern (@Grummz), 1,814 likes and 87,000 views in the first hour
Chatgpt is Down. Grok is Down. Cursor is Down. Anthropic Claude is Down. Vibecoders and Solo founders rn:
Ariska (@ariskaa_ai), 349 likes, 36,000 views
It looks like a national state brought down Claude, ChatGPT, and Grok.
Charles Hoskinson (@IOHK_Charles), 847 likes, 136,000 views; no company has suggested an attack
No company has suggested an attack. Hoskinson's post offers no evidence for one.
Who felt it in Australia
The outage began after the end of the Australian working day. Night-shift staff, offshore teams and automated jobs that call the APIs overnight were the users affected here, and developers running Claude Code or Cursor lost sessions part-way through.[5][4] Bushletter's audit of 136 Australian agencies, published on Thursday, found 118 selling an AI-search service on the same APIs.[10] OpenAI closed its incident at 2:55am Sydney time: "The issue causing elevated errors across ChatGPT and Codex has been resolved. Some Codex remote control users may need to pair their mobile device again." OpenAI, Anthropic and xAI are expected to publish post-incident reports in the coming days.
SOURCES & CITATIONS
- OpenAI status page: elevated errors across ChatGPT and Codex, 3 September 2026
- Is OpenAI's ChatGPT down? Thousands of users report outages among Grok, Gemini, Claude, Newsweek, 3 September 2026
- Claude status page: elevated errors for multiple models, 3 September 2026
- All the major AI chatbots are experiencing outages right now, Gizmodo, 3 September 2026
- It's not just you; ChatGPT, Claude, and Grok are all down in confirmed outages, 9to5Google, 3 September 2026
- Azure status page, 3 September 2026
- Cloudflare status page, 3 September 2026
- ChatGPT, Gemini, Claude face widespread outages, The Hindu, 3 September 2026
- Anthropic confirms Claude is down, multiple models affected, BleepingComputer, 3 September 2026
- Most Australian agencies now sell AI search, and only a couple deliver, Bushletter, 3 September 2026
FREQUENTLY ASKED QUESTIONS
Is ChatGPT back up?
Was it a Cloudflare or Azure outage?
What should a business that depends on these APIs do?
Alex Mercer writes about technology, energy and infrastructure. He likes the physical end of the story: the plants, the grids and the machines that everything else depends on.
