OpenAI outage history

A small subset of OpenAI users experienced authentication errors when using the Codex API, impacting Codex API usage. Some reported issues included Codex getting stuck on the thinking stage and being logged out when viewing task histories. The impact began around 5:01 PM UTC and, after a mitigation was applied, monitoring continued, with issues stabilizing by 5:56 PM UTC. OpenAI noted the elevated Codex API authentication errors in status updates and indicated that mitigation was in place and the system was being watched. A two-step response—mitigation deployment 5:16 PM UTC and eventual restoration—led to full recovery by the end of the window.

OpenAI's ChatGPT experienced elevated errors affecting Thinking mode, with some users encountering failures during conversations. The incident began at 1:57 PM UTC when OpenAI's status page listed a degraded condition affecting ChatGPT Conversations and ChatGPT Work. The problem expanded to include Image Generation by 2:11 PM UTC and was consolidated to Thinking mode and Image Generation by 2:16 PM UTC.

Mitigations were implemented at 2:57 PM UTC, and operators reported improvements as fixes rolled out. By 4:56 PM UTC the affected components — Conversations, Image Generation, and ChatGPT Work — had returned to normal, marking the incident as resolved.

OpenAI's ChatGPT login and chat services were unavailable for many users globally, preventing signups and logins on chatgpt.com and blocking access to chat features. Some requests returned 502 Bad Gateway errors, and users could not start new chats or view past conversations. The issue appeared in user reports around 11:45 PM UTC and was acknowledged by OpenAI around 12:06 AM UTC with mitigation efforts underway.

OpenAI's ChatGPT.com experienced a global outage affecting signups, sign-ins, and access to chats and history. Users across regions reported being unable to sign in, load chats, or view past conversations, with 502 gateway errors observed on login. The incident began at Aug 19, 2026 11:58 PM UTC, and StatusGator detected the issue via an Early Warning Signal before OpenAI acknowledged it.

OpenAI acknowledged the problem around 12:06 AM UTC and said mitigation was being implemented, with monitoring continuing and additional steps by 12:26 AM UTC. The ChatGPT / Login component returned to an up state by 12:56 AM UTC, with full recovery observed by Aug 20, 2026 1:00 AM UTC. The outage appeared global, reflected by user reports from the United States, Europe, Asia, and other regions indicating widespread login and chat access issues.

OpenAI's ChatGPT / Sites deployments experienced elevated errors, causing some deployments to fail or run slowly. User reports began around 6:45 PM UTC with a second at 7:00 PM UTC, indicating the issue was impacting the Site deployment workflow for a portion of users. The impact appeared centered on the Site deployment process rather than the broader platform, with affected users seeing degraded deployment reliability.

OpenAI reported an issue impacting a subset of GPT-5.6 Terra users, resulting in intermittent incomplete or malformed responses and occasional unresponsive chats. The impact began around 10:26 PM UTC and affected core ChatGPT experiences including conversations and prompt processing. OpenAI initiated mitigation efforts to address the problem as engineers investigated the root cause and worked toward a fix. The issues appeared across the ChatGPT / Conversations component as well as APIs / Chat Completions and APIs / Responses, with users reporting delays, loading problems, and unexpected errors.

Over the next several hours, status updates indicated progress toward restoration, and services gradually returned to normal. The incident was resolved when all listed components returned to a healthy state by 8:52 AM UTC.

OpenAI experienced a degradation impacting ChatGPT Conversations, with Free-tier users reporting elevated errors beginning around 3:22 PM UTC. An additional issue affected Enterprise users trying to create custom RBAC roles. OpenAI attributed the disruption to a downstream provider problem and rolled out mitigations while monitoring recovery. The incident produced a noticeable stream of user reports across several regions, reflecting a degraded experience for a portion of users as conversations error’ed or failed under load. Mitigations were applied over the next several hours and monitoring continued, with updates indicating progress on recovery around 4:56 PM UTC and 9:26 PM UTC and a downstream-provider note at 10:11 PM UTC.

OpenAI faced degraded ChatGPT conversations for Free users starting in the afternoon of Aug 13, 2026 3:22 PM UTC after an initial issue with creating custom RBAC roles for Enterprise users. The earliest status update indicated an error during RBAC role creation at Aug 13, 2026 2:07 PM UTC, and by Aug 13, 2026 3:22 PM UTC the problem had expanded to user-facing chat experiences for Free users.

OpenAI implemented mitigations over the evening and into the night. Updates showed mitigation was applied and monitoring ongoing by Aug 13, 2026 4:56 PM UTC, with continued issues and downstream provider involvement discussed in later notices. Recovery for the ChatGPT / Conversations component followed a mitigation track and was reported as returned to normal by Aug 14, 2026 12:01 AM UTC.

OpenAI users faced elevated errors across the API, Codex, and Work Mode beginning around 9:36 PM UTC. The OpenAI status page signaled a degraded state as engineers worked on a mitigation, while customers reported slow responses, timeouts, and failures when processing requests. Examples from user reports included streams disconnecting before completion and model usage hitting capacity, with many users unable to complete typical tasks.

Mitigation efforts were put in place around 10:17 PM UTC, and recovery progressed over the next hours. By 10:46 PM UTC the status page had moved back to up, indicating services had recovered. The incident generated a noticeable uptick in user-reported issues across regions, reflecting a broad impact on API calls, Codex interactions, and Work Mode features during the incident.

OpenAI ChatGPT users began experiencing elevated error rates around 12:57 PM UTC, with issues affecting ChatGPT Conversations and ChatGPT Work. Prompt submissions could fail, conversations could stall, and streaming sessions could disconnect, according to user reports. A subset of users also encountered problems with ChatGPT Go conversations. The disruption persisted for roughly two hours, impacting a broad slice of daily use.

OpenAI published updates indicating ongoing investigation and mitigation. By 2:36 PM UTC the team had applied mitigation and was monitoring recovery, and services began returning to normal by 3:02 PM UTC for ChatGPT Conversations and ChatGPT Work. User reports tapered off as stability returned.