OpenAI logged two separate incidents on 13 November 2026: a brief publishing error on its Sites feature, and a much longer delay in Compliance API cost data that ran for 56 hours 7 minutes. Both were OpenAI-side faults, confirmed and tracked on OpenAI's own status page.
Started
03:59 UTC
Duration
Lasted 1h 59m
Source
IsDown
Next time
OpenAI is degraded. Is your product working?
An incident on their side does not always mean an outage on yours, and the only way to know is to already be watching your own endpoints. Set that up in under a minute and you will be notified the next time your own site is affected. And there will be a next time.
50 monitors for your own websites. No card.
Affected components: Compliance, APIs
What happened?
The day's first incident started at 03:59 UTC, when OpenAI identified elevated errors affecting site publishing. That entry carries only one update, posted at the moment OpenAI said it was working on a mitigation, and the incident closed after 1 hour 59 minutes with no further detail published on what the mitigation was or why the errors occurred. The second incident began at 16:10 UTC and lasted far longer, 56 hours 7 minutes in total. OpenAI's own description was specific: delays in publishing Organization cost data through the Costs API, with API and ChatGPT usage explicitly called out as unaffected. That distinction matters, because it tells you this was a reporting and billing visibility problem, not a request-serving one. OpenAI posted three updates across the two days. At 18:56 UTC on the first day it said a mitigation had been applied and that it was monitoring for recovery, while warning that cost data delays were still expected. A further monitoring update followed at 05:31 UTC with the same caveat repeated almost word for word. No update explains what caused the delay in the first place, and no closing note describing root cause has been published. Two components were named across the day: Compliance and APIs. No user reports were filed during either incident's window, so the record here is built entirely from OpenAI's own status page language, not from anyone independently confirming the symptoms.
Learning
This outage sat in a part of the stack that most uptime checks never touch: billing and cost reporting data, not the API responses your product actually calls. If you were polling OpenAI's chat or completion endpoints during either incident, you'd have seen nothing wrong, and you'd have been right, because OpenAI itself said API and ChatGPT usage were not affected. The failure was in a downstream reporting pipeline that surfaces organization cost data, a layer that sits behind the API rather than inside the request path you monitor. That's a useful shape to recognise: a provider can be fully responsive to your traffic while one of its internal services, like cost aggregation, silently falls behind by days. Watching your own endpoints only tells you whether requests succeed, not whether the provider's supporting systems, dashboards, or administrative data are current. Your monitoring covers your half, UptimeRobot now watches the provider's half too, so next time you know which side broke without guessing.