Disruption at 22:45 UTC — 40 reports over 37m. Most common: 'Not working' (64%).
No problems reported at OpenAI right now
Problem reports for OpenAI are at their usual level right now. The most recent incident was about 7 days ago.
reports · 24 h · /15 min
1h2↑15m0↓5m–
OpenAI not working for you?
what people report · last 24 h
Service25%
App25%
Router25%
Not working25%
where from · last 24 h
no located reports in this window
outage map · last 24 h
live pulse
- social
Is dots down? @OpenAI5h - social
@OpenAI @sama Doesn’t explain the modem going down tho. 👀7h - social
Codex can neither work properly with blockchain products nor mac/window/android applicati…8h - report🇨🇦Not working - Lindsay2d
- report🇺🇸Problem report - Newalla17d
- report🇺🇸Not working - San Antonio17d
catalogOpenAI products: ChatGPT.
OpenAI outage record
all services by year →4 incidents in the past 12 months, none major.
Since 2025, down.now has recorded 4 outage days on OpenAI; the largest was Sep 3, 2026 with 138 reports, across 5 countries.
Outage — problem reports peaked at 44/15min around 15:05 UTC; 138 reports over 1h 7m. Most common: 'Not working' (74%); reports from the United States, Canada, Japan and 2 more countries. The company acknowledged it 13m after the first reports.
- reports
- 138
- duration
- 1h 7m
- top problem
- Not working 74%
- most affected
- US 33% · of 5 countries
- peak
- 44 / 15 min @ 15:00 UTC
- official ack
- +13m after reports
Went down at the same time as Claude, ChatGPT and Gemini.
story · 6 beats
- 14:46 Problem reports started climbing at 14:46 UTC.
- 14:58 The company acknowledged it at 14:58 UTC — 13m after the first reports.
- 15:05 Peak at 15:05 UTC: 44/15min.
- 15:45 Down to under half the peak by 15:45 UTC.
- 15:53 Problem reports faded around 15:53 UTC.
- 16:56 The company gave the all-clear at 16:56 UTC.
reports · 2026-09-03 · /15 min
incident timeline
- 14:46Problem reports started climbing.
- 14:58The company acknowledged it — 13m after the first reports.
- 15:05Peak: 44/15min.
- 15:45Down to under half the peak.
- 15:53Problem reports faded.
- 16:56The company gave the all-clear.
outage map
Outage — problem reports peaked at 23/15min around 09:35 UTC; 90 reports over 1h 29m. Most common: 'Not working' (64%); reports from the United Kingdom, Indonesia, Israel and 10 more countries. The company acknowledged it 15m after the first reports.
- reports
- 90
- duration
- 1h 29m
- top problem
- Not working 64%
- most affected
- GB 18% · of 13 countries
- peak
- 23 / 15 min @ 09:15 UTC
- official ack
- +15m after reports
story · 6 beats
- 09:04 Problem reports started climbing at 09:04 UTC.
- 09:18 The company acknowledged it at 09:18 UTC — 15m after the first reports.
- 09:35 Peak at 09:35 UTC: 23/15min.
- 10:15 Down to under half the peak by 10:15 UTC.
- 10:33 Problem reports faded around 10:33 UTC.
- 11:57 The company gave the all-clear at 11:57 UTC.
Incident at 13:53 UTC — 48 reports over 1h 4m.
OpenAI comments
anonymous · no accountcommon questions
Is OpenAI down right now?
Problem reports for OpenAI are at their usual level right now. The most recent incident was about 7 days ago.
Is OpenAI down for everyone or just me?
Right now down.now isn't seeing a broad OpenAI outage, so the problem is most likely local to you — your connection, your device, or the app itself. Try the app on mobile data instead of Wi-Fi, then restart the router. If reports start arriving from other people, this page flips to an outage within minutes.
When did OpenAI last go down?
The last recorded OpenAI incident was about 7 days ago, on 2026-09-25. The outage record below has the peak time, duration and what people reported.
How can I report a problem with OpenAI?
Use the report panel on this page and pick what's not working — it takes one tap and no account. Reports are anonymous, and they feed the live signal on this page immediately.
How does down.now know if OpenAI is down?
Down.Now watches two crowd signals for OpenAI — anonymous reports submitted here, and public social-media posts about it — and compares the live volume against what's normal for that hour of the day. A sustained spike above that baseline opens an incident, which is usually visible before an official confirmation. We measure what users report, not OpenAI's servers directly.