
OpenAI Confirms Global ChatGPT and API Outage Amid Login Failures
OpenAI has confirmed a major global outage affecting ChatGPT, its API, and the Codex coding platform. The disruption began at approximately 8:00 PM Eastern Time on Wednesday, August 19, leaving users worldwide unable to log in, sign up, or send messages to the artificial intelligence chatbot. Users attempting to access ChatGPT are encountering a frozen interface, with the sidebar stuck on a continuous loading animation. Any attempts to prompt the chatbot result in a "too many concurrent requests" error message. The login and registration portals on chatgpt.com are also completely down, preventing new or returning users from authenticating their accounts. Beyond the consumer-facing web interface, the outage has heavily impacted developers who rely on OpenAI's infrastructure. Up to 12 endpoints for the OpenAI API are currently listed as experiencing issues on the company's official status page. Codex, the AI model that powers various coding assistance tools, is also caught in the service disruption. The company has officially acknowledged the situation. By 8:15 PM ET—roughly 14 minutes after the problems began—OpenAI updated its system status page to indicate that the root cause of the widespread login issues had been identified. The engineering team stated they are actively "working on implementing a mitigation" to restore access across all affected services. Despite the prompt identification of the problem, the outage remains ongoing, and independent tests confirm that services are still unreachable. The disruption affects a broad base of users globally, with significant connectivity failures already documented across the United States and Europe. University students in Mogadishu and software developers across East Africa relying on ChatGPT or the OpenAI API will find these tools completely unresponsive until servers are restored. Local tech startups and businesses that have integrated OpenAI's infrastructure for automated customer service will also return connection errors to users until the company deploys its fix.
Read our sourcing and corrections policy
This article was prepared by our automated editorial system, which summarized and translated the source above. No human editor reviewed this article individually before publication.



Be the first to comment on this story!