What happened when ChatGPT went down on August 31, 2026
On August 31, 2026, ChatGPT had its biggest outage of the season — searches for “chatgpt down” spiked to roughly ten times any other AI outage in the following three months. Here is exactly what happened, minute by minute, rebuilt from OpenAI’s own status updates.
The timeline (all times UTC)
- 15:04IdentifiedElevated latency across the impacted services. OpenAI starts working on a mitigation.
- 15:21IdentifiedNow elevated errors too. ChatGPT Work mode is unavailable, with Plus users particularly impacted.
- 16:02OngoingUsers across multiple subscription plans are unable to start or continue tasks in ChatGPT Work.
- 17:01–19:29OngoingWork tasks keep failing through four more status updates over the next ~2.5 hours.
- 20:01MonitoringA mitigation is applied; OpenAI monitors recovery.
- 20:28ResolvedAll impacted services have fully recovered.
The core incident ran five hours and twenty-four minutes, from 15:04 to 20:28 UTC. The pattern is a familiar one: latency first, then outright errors, and the paid Work (business) tier took the worst of it — exactly when teams relying on it for actual work were online.
It didn’t end there
That resolution wasn’t the end of it. Over the following days OpenAI posted a string of further incidents — a rough patch, not a one-off:
The takeaway
Even the most-used AI in the world goes down — and when it does, it’s the paid, work-critical tiers that feel it first. The only way to know whether it’s you or them is to check the source. We mirror ChatGPT’s official status in plain English on the ChatGPT status page, track every provider’s outages in the outage history, and rank who breaks least on the reliability board.