On April 20, 2026, OpenAI experienced a significant and widespread service disruption that disabled access to its primary artificial intelligence platforms for several hours. The outage affected ChatGPT, the Codex programming model, and the OpenAI API Platform, causing a complete halt in services for millions of individual users and thousands of enterprise customers worldwide. According to the official OpenAI status dashboard, the incident began at approximately 07:15 UTC, with reports of connectivity issues peaking within the first thirty minutes of the event as traffic flooded support channels.
Technical logs indicated that users attempting to access ChatGPT via web and mobile interfaces were met with 503 Service Unavailable and 504 Gateway Timeout errors. The disruption extended to the API v5 and v4 endpoints, which are critical for third-party applications and automated workflows. Developers reported that calls to the gpt-5-turbo and gpt-4o models were failing consistently, returning internal server errors rather than the expected responses. OpenAI confirmed that the outage was global in scope, with the highest concentration of reports originating from North America, Western Europe, and parts of East Asia.
OpenAI issued its first official statement at 07:45 UTC, acknowledging that the company was investigating an issue resulting in high error rates across all services. By 09:30 UTC, the company identified the root cause as a failure in the primary database cluster responsible for session management and authentication. This failure prevented the system from validating user credentials and maintaining active chat sessions. Technical teams initiated a failover to redundant systems, but the process was delayed by a secondary bottleneck in the load-balancing layer that manages traffic across regional data centers.
The outage impacted an estimated 200 million daily active users who rely on the platform for various professional and personal tasks. Enterprise clients, including those utilizing the ChatGPT Enterprise and Team tiers, reported significant disruptions to internal operations. For organizations integrated with the Codex API, the outage halted automated code generation and real-time debugging tools. OpenAI reported that while data integrity remained uncompromised, the total downtime lasted approximately four hours and twelve minutes before partial service was restored at 11:27 UTC.
As of 14:00 UTC on April 20, OpenAI stated that services had returned to operational status for the majority of users, though some might still experience intermittent latency as the system re-indexed cached data. The company has committed to providing a full post-mortem report within 48 hours to detail the infrastructure vulnerabilities that led to the collapse. This event marks the longest period of total downtime for OpenAI's ecosystem since the previous year's infrastructure upgrade, highlighting ongoing challenges in scaling global AI services.