OpenAI confirmed a partial ChatGPT outage on August 31, 2026, with Work mode unavailable to paying Plus subscribers. Downdetector logged more than 2,000 user reports by 9:01 a.m. PT.
Key Takeaways
- OpenAI listed ChatGPT as a partial outage and said Work mode was unavailable to Plus subscribers.
- Downdetector recorded more than 2,000 problem reports by 9:01 a.m. PT on August 31, 2026.
- The status page moved from a latency warning to an error warning in about 17 minutes.
- OpenAI acknowledged the errors and said it was working on a mitigation, with no restore time given.
What We Know?
OpenAI’s incident page opened with a latency notice and upgraded about 17 minutes later to elevated errors. The second update named the group hit hardest: “Plus users are particularly impacted, where Work mode is currently unavailable.” A later update widened the scope to users across multiple subscription plans who could not start or continue tasks in ChatGPT Work.
Downdetector, the outage tracker, counted more than 2,000 by 9:01 a.m. PT. That figure tracks complaints, not accounts, so it sets a floor rather than a headcount.
User reports indicate problems with OpenAI since 11:25 AM EDT.
— Downdetector (@downdetector) August 31, 2026
How is it affecting you? #OpenaiDownhttps://t.co/qU4YVGAOdp
What We Don’t Know?
OpenAI’s updates describe symptoms and say nothing about cause. Open questions:
- What triggered the errors, and whether the trigger sits inside ChatGPT Work or in shared infrastructure?
- How many accounts lost access, and across which plans?
- Whether tasks in progress were lost or recoverable?
- When OpenAI expects a mitigation to land?
Anyone mid-task in ChatGPT Work should copy drafts into local storage before retrying, which helps reduce the risk of lost output.
SQ Magazine’s Takeaway
The revealing detail is which tier broke. Work mode is the paid surface aimed at business users, and it failed while the rest of ChatGPT stayed partly usable. That may point to the agentic task layer rather than the model serving path, though OpenAI has confirmed neither. For a customer who moved daily workflows onto a platform now running at consumer scale, losing that component is a full outage.
What’s next is narrow. IsDown puts the median resolution time for ChatGPT incidents at 2 hours and 28 minutes across 208 incidents since October 2025, so teams with deadlines should plan for the slower end. The signal worth watching is whether OpenAI’s follow-up names a cause or only reports recovery, because an incident closed without explanation leaves business customers unable to judge whether it returns.
