OpenAI Outage History

OpenAI’s ChatGPT paid plan users experienced elevated errors when sending and receiving messages, with some reporting streaming failures and unresponsive chats. The disruption began around May 21, 2026 7:35 PM UTC, and StatusGator detected the issue about 46 minutes before OpenAI formally acknowledged it at May 21, 2026 8:21 PM UTC. The early warning signal ended around May 21, 2026 8:13 PM UTC as initial user reports subsided. OpenAI issued a mitigation and monitored recovery, noting ChatGPT / Conversations as the affected component.

Mitigation progressed through the evening and the incident began to ease, with OpenAI updating the status to up for ChatGPT / Conversations by May 21, 2026 9:52 PM UTC. The issue appeared to be global in scope, impacting the messaging experience for paid-plan users. By the time recovery was complete, most users experienced restored responses, though some sporadic issues were reported during the gradual return to normal.

From around 2:00 PM UTC, OpenAI users reported issues in chat and completion workflows, with error messages, stalled message streams, and prompts failing to generate responses. The disruption affected real-time conversations and prompt-based tasks across multiple regions, leaving some paid and free users unable to get timely results. StatusGator detected the issue at May 21, 2026 2:27 PM UTC, with activity continuing until 2:53 PM UTC, indicating a roughly fifty-minute period of degraded service.

Users on OpenAI's ChatGPT paid plans experienced elevated errors affecting the ChatGPT / Conversations component. Reports described error messages, stalled message streams, and prompts failing to generate, degrading the experience for premium users across multiple regions. The issue began at 1:56 PM UTC as OpenAI flagged elevated error rates and initiated a mitigation.

OpenAI reported elevated error rates for GPT-5.4 and GPT-5.5, affecting the API endpoints used for Chat Completions and Responses. The issue began at 11:36 PM UTC and caused degraded experience for users making model requests. OpenAI implemented mitigation and continued monitoring, with recovery progressing over the next hour. By 12:02 AM UTC the affected endpoints had started to recover, and all services were fully restored by 12:41 AM UTC.

OpenAI reports that GPT-5.5 experienced degraded performance beginning May 15, 2026 4:11 PM UTC. Users encountered slower responses and intermittent issues across GPT-5.5-enabled experiences, including ChatGPT interactions and Codex services. The problem persisted for roughly 32 hours as OpenAI investigated the cause and implemented mitigations, with notable improvements after 12:37 AM UTC and a return to normal operation by May 17, 2026 12:16 AM UTC.

OpenAI provided status updates acknowledging the investigation and eventual mitigation. User reports during the period described symptoms such as loading delays, chats not loading or responding, and Codex-related tasks intermittently failing, indicating a broad, but not complete, degradation across multiple features and services.

Developers using OpenAI Codex Cloud and Code Review faced elevated errors beginning at May 14, 2026 9:01 PM UTC. OpenAI's status updates indicated these components were experiencing a high failure rate and that a mitigation was being implemented as teams investigated the root cause.

The disruption persisted for about 30 minutes while the company worked on a fix, and the services returned to normal when the status changed to up at 9:31 PM UTC. The impact was confined to Codex Cloud and Code Review, with no broader OpenAI services indicated as affected.

Users attempting to use ChatGPT without signing in encountered errors and pages failing to load. The outage began at 11:07 PM UTC when the ChatGPT / Login component went down for logged-out users. A mitigation was applied and monitoring started at 11:21 PM UTC, and access began returning to normal by 11:32 PM UTC. The disruption lasted about 25 minutes and affected unauthenticated user sessions across multiple regions, with reports coming from the United States and other locations describing errors loading ChatGPT pages.

An outage in OpenAI's Realtime API SIP flow began at 6:31 AM UTC, causing a disruption across Codex 5.5 usage. The SIP/WebRTC flow down condition led to degraded performance and higher error rates for users attempting to use Codex interfaces, with multiple Codex components affected—Codex API, Codex Web, CLI, the VS Code extension, and the Codex App. The first explicit signals appeared around 8:21 AM UTC as Codex 5.5 engines entered a high error-rate state, and by 9:01 AM UTC free-user requests were experiencing elevated errors and degraded latency across Codex services.

Users across multiple regions experienced authentication and login failures as OpenAI flagged elevated error rates tied to GPT 5.5. The issue affected authentication flows, with reports describing identity_provider_mismatch errors during sign-in attempts. Impact began around 5:07 PM UTC as OpenAI announced mitigation and monitoring of recovery, and issues persisted into the evening as user reports accumulated.

OpenAI stated that mitigation had been applied and recovery was being monitored, with the service largely returning to normal by 8:16 PM UTC. Reports came from North America (including the United States and Canada), indicating a broad but partial outage affecting authentication rather than a full platform outage.

An outage impacted file uploads in ChatGPT and Codex Cloud task creation. Users reported errors uploading documents, with PDFs and other files failing or disappearing during the process. The disruption began at 1:26 AM UTC and lasted for about four hours, affecting ChatGPT / File uploads and Codex / Codex Web across regions. The disruption was evident in multiple regional reports describing upload failures and interrupted workflows.

OpenAI stated they were investigating the issue and applying mitigations. By 3:26 AM UTC they had applied the mitigation and were monitoring the recovery, and by 5:37 AM UTC services had returned to normal.