OpenAI reported a widespread service disruption on April 20, 2026, that rendered its ChatGPT platform and developer tools inaccessible to thousands of users worldwide. The incident began at approximately 14:15 UTC, with reports of connectivity issues peaking within twenty minutes. According to data from network monitoring services, the outage affected the ChatGPT web interface, mobile applications, the Codex programming model, and the enterprise API platform.
During the height of the disruption, users attempting to access the service were met with 503 Service Unavailable and 502 Bad Gateway errors. Internal monitoring indicated that the API response latency spiked from a standard average of 200 milliseconds to over 15 seconds before failing entirely. OpenAI’s official status page confirmed the incident at 14:32 UTC, stating that the company was investigating elevated error rates across all model endpoints.
The outage impacted a broad range of services, including ChatGPT Plus subscribers and enterprise clients who rely on the API for integrated applications. DownDetector, a platform that tracks service interruptions, recorded over 15,000 individual reports from users in North America, Europe, and Asia within the first thirty minutes of the event. The disruption also extended to the Codex engine, which powers various third-party coding assistants, leading to temporary workflow interruptions for software developers.
At 15:10 UTC, OpenAI updated its status report to indicate that a fix had been implemented and services were beginning to recover. The company identified the root cause as a failure in a primary database cluster responsible for managing session tokens and authentication requests. This failure triggered a cascading effect that overwhelmed the load balancers, preventing new connections from being established. By 15:25 UTC, OpenAI reported that ChatGPT and the API platform were fully operational, though some users continued to experience intermittent latency for an additional thirty minutes as traffic stabilized.
In a brief technical update released following the restoration, OpenAI engineers noted that the infrastructure failure occurred during a routine maintenance window intended to scale capacity for the GPT-5 model architecture. The company stated it is conducting a full post-mortem to prevent recurrence and will provide a detailed incident report to its enterprise partners. This event marks the third significant disruption to OpenAI’s infrastructure in the current calendar year, following a similar two-hour outage recorded in February.