OpenAI Outage History

OpenAI ChatGPT file uploads experienced elevated errors beginning at Sep 22, 2025 4:21 AM UTC, with users reporting difficulties uploading files to ChatGPT. This degraded the file upload functionality for anyone trying to attach documents or media to conversations. The incident lasted until Sep 22, 2025 9:56 AM UTC, a duration of about 5 hours and 35 minutes, during which the ChatGPT / File uploads component was identified as impacted and mitigation efforts were underway.

OpenAI posted 2 status updates during the incident to communicate progress and mitigation. The updates described ongoing efforts and, by the end of the incident, restoration of normal file upload functionality.

The OpenAI ChatGPT service experienced a partial outage affecting the ability to create new projects across four components: ChatGPT Web, ChatGPT Mobile for iOS, ChatGPT Mobile for Android, and ChatGPT File uploads. Users attempting to create projects would have encountered failures or errors. The incident began at 7:56 PM UTC UTC and was resolved by 11:12 PM UTC UTC, lasting approximately 3 hours and 16 minutes.

OpenAI's ChatGPT file uploads feature experienced an outage that prevented users from uploading files to the impacted services. Users reported errors such as “Something went wrong” and difficulties opening new projects, with observations from multiple regions including the United States, Poland, and Nigeria, indicating a broad impact on the file-upload workflow. The incident began with a status change to warn at Sep 20, 2025 1:41 AM UTC and was resolved by Sep 20, 2025 6:57 PM UTC, lasting about 17 hours and 16 minutes.

OpenAI issued two official status updates during the incident, noting that the issue had been identified and mitigations were underway, and later confirming restoration once the service returned to normal at Sep 20, 2025 6:57 PM UTC. There was no Early Warning Signal detected for this incident, and the disruption affected the ChatGPT / File uploads component.

With the service restored, users reported normal file upload functionality resumed. This incident highlights the impact a single component can have on workflows that rely on file uploads within ChatGPT.

OpenAI ChatGPT components experienced elevated error rates starting at 4:51 PM UTC, affecting ChatGPT / Agent and later ChatGPT / Conversations. User reports around 6:05 PM UTC described issues uploading or deleting project files. The incident lasted about 1 hour 40 minutes, with services returning to normal after 6:31 PM UTC. OpenAI issued three status updates during the incident to communicate investigation progress and recovery, classifying the event as partial degradation rather than a full outage.

OpenAI reported elevated errors affecting the Conversations with Images feature for Free Users. The outage began at 10:07 PM UTC and was resolved by 10:41 PM UTC, lasting about 34 minutes. OpenAI posted two status updates during the incident, noting elevated errors and that mitigation was underway before confirming the issue was resolved.

OpenAI reported elevated errors affecting login flows and GPT services, impacting authentication and chat generation. Affected components included ChatGPT / GPTs, Sora / Login, APIs / Login, and ChatGPT / Login. Users reported sign-in issues, slow or stalled responses, and occasional errors in replies or image generation. The incident began at Sep 16, 2025 3:46 PM UTC and progressed with partial recoveries, with login components returning to normal by Sep 16, 2025 5:16 PM UTC and GPTs restored by Sep 17, 2025 5:36 PM UTC, lasting roughly 26 hours overall. The issue appeared to have global reach, with user reports from multiple regions.

OpenAI's API platforms experienced a multi-day outage beginning at 1:11 AM UTC and not fully resolved until 4:51 AM UTC. Early in the incident, OpenAI flagged impact to file_search retrieval quality across Assistants API, Responses API, and Retrieval API, with users reporting degraded behavior and intermittent errors across related endpoints.

The OpenAI status page issued multiple updates and mitigations over the course of the incident. A mitigation was applied around 1:06 AM UTC and recovery was monitored thereafter. Reports of elevated errors and rate‑limit issues, including GPT-4o and agent-related tasks, followed as the outage expanded to a broader set of services, culminating in elevated error rates across the API Platform on 11:46 AM UTC affecting Embeddings, Batch, Files, Images, Audio, Realtime, Login, Fine-tuning, Completions (legacy), Chat, Compliance, Moderations, and Uploads. Recovery progressed in the days that followed and components returned to normal by 4:51 AM UTC.

The incident affected a broad set of OpenAI APIs and services, with a noticeable volume of user reports indicating widespread impact across Embeddings, Batch, Files, Images, Audio, Realtime, Login, Fine-tuning, Completions (legacy), Chat, Compliance, Moderations, Uploads, and related assistant and research interfaces. The outage lasted roughly four days before full recovery.

OpenAI Batch API experienced a brief degradation as calls to the Batch API returned elevated error rates. The incident began at Sep 11, 2025 12:46 AM UTC and was resolved by Sep 11, 2025 12:51 AM UTC. OpenAI issued two status updates during the window, noted elevated errors and a mitigation, and the Batch API returned to normal after the incident. The impact was isolated to the Batch API component.

OpenAI reported a brief incident during which ChatGPT / Agent services experienced an elevated error rate, which could cause some user requests to fail or respond with errors. The OpenAI status page indicated that users were experiencing elevated errors for the impacted services and that a mitigation was being implemented to restore normal operation.

The Web Search tools within the OpenAI Responses API experienced elevated errors, impacting users attempting to fetch web results via the API. The incident started at Sep 5, 2025 9:37 AM UTC and lasted about 14 minutes, with services returning to normal by Sep 5, 2025 9:51 AM UTC.

OpenAI issued 2 status updates during the incident, confirming the issue and later the mitigation and restoration of normal operations. Because the disruption was brief, the overall user impact was limited to a short period of degraded performance.