OpenAI outage hits globally

OpenAI users around the world lost access to ChatGPT, Codex, and the company’s application programming interface platform on Monday, April 20. (status.openai.com) OpenAI’s status page said affected users were “unable to access ChatGPT, Codex and API Platform” during a partial outage that began at 8:13 a.m. and was marked resolved at 11:48 a.m. Eastern time. The company posted “monitoring” updates at 10:43 a.m. and 11:17 a.m. after saying it had applied a mitigation. (status.openai.com) Reports from users outside OpenAI’s own systems pointed to a broad disruption rather than a single-region failure. Yahoo Tech, citing Downdetector data, reported a spike starting around 10:05 a.m. Eastern, with more than 8,700 reports in the United Kingdom and about 1,900 in the United States at the peak. (tech.yahoo.com) The outage crossed OpenAI’s consumer and developer products at the same time. The incident page listed 12 affected ChatGPT components, and OpenAI’s history feed tied the same event to the Responses service on the application programming interface side. (status.openai.com 1) (status.openai.com 2) That overlap matters because ChatGPT’s chat window, Codex coding tools, and the application programming interface are separate front ends for shared back-end systems. When one of those common layers fails, the break can show up at once for consumers, software developers, and companies that build on OpenAI models. (status.openai.com) (openai.com) OpenAI’s public status dashboard also shows how central uptime has become to the company’s pitch to users and customers. As of April 21, the dashboard listed aggregate uptime of 99.99% for application programming interfaces, 99.85% for ChatGPT, and 100% for Codex for the January-to-April 2026 window, while noting that individual customer availability can vary by tier and feature. (status.openai.com) The company has logged other service problems in recent months, including a Codex incident on March 10 and earlier application programming interface login and authentication issues in the history feed. Monday’s outage stood out because a single incident covered ChatGPT, Codex, and the application programming interface together. (status.openai.com 1) (status.openai.com 2) By late Monday, OpenAI’s incident page said “all impacted services have now fully recovered.” For users who now treat ChatGPT like search, Codex like a coding assistant, and the application programming interface like infrastructure, the interruption was a reminder that all three can still go dark at once. (status.openai.com)