OpenAI's public status page said Sunday that ChatGPT and related systems were fully operational, with no active incidents flagged. That quiet reading matters because September has been noisy for the chatbot's uptime record. A same-week review of OpenAI's own status history counted 37 separate incident entries from Sept. 1 through Sept. 26, logged on 21 of those 26 days.

Most of those entries were short. Our Nation Online timed four incidents in detail — including a Sept. 26 Codex disruption and the larger Sept. 3 ChatGPT-and-Codex event — and each resolved in under about two and a half hours on the status log. OpenAI's own write-up on the Sept. 3 routing error said the main user-facing hit was mitigated in roughly 37 minutes after problems began around 7:43 a.m. Pacific, though outside trackers still logged tens of thousands of Downdetector-style reports that day. The company's trailing ChatGPT uptime on the status page sat near 99.7% as of Sunday's check.

That pattern helps explain why status-page and "is ChatGPT down" searches keep flaring even when there is no long weekend outage. Users hit an error, open status.openai.com or an outage tracker, and search interest jumps. CNBC's Aug. 31 coverage of a ChatGPT Work elevated-errors window — roughly three hours, with more than 1,100 Downdetector reports at one afternoon snapshot — showed the same cycle for OpenAI's enterprise agent product. OpenAI's history page also lists repeated Work, Plus/Pro, Codex, and regional error entries through mid-to-late September, nearly all marked recovered by Sunday.

OpenAI Status and the September incident tally agree the public record is a string of brief flaps, not a multi-day Sunday blackout. They still leave open how often account-specific problems never appear on the aggregate status board — Our Nation Online flagged a multi-day Codex "at capacity" community thread that did not clearly match one long official incident — and how large the next Downdetector spike will look when another routing or Work-mode error lands.