OpenAI’s conversational artificial intelligence service ChatGPT suffered a global access outage on the morning of July 15. Core functions, including login and chat generation, were paralyzed, causing significant disruption for users.
OpenAI announced earlier that morning via its official status page that it was aware of the issue and had launched an investigation. The company stated it is “investigating login issues and intermittent errors affecting ChatGPT.”
The outage began to be widely reported around 8:30 a.m. Korean time. According to DownDetector, a website that tracks internet service disruptions, reports of ChatGPT access failures surged around 8:32 a.m., with 1,395 error reports filed during that period.
Users attempting to access ChatGPT via web browsers were met with messages such as “upstream connect error” or “reset/connection closed before headers,” rendering the service unusable.
The scope of the disruption disclosed by OpenAI was extensive. Errors were confirmed across a total of 15 features, including chat generation, login, ChatGPT tasks, search, file uploads, and image generation.
However, the outage appears to have been confined to the ChatGPT service. OpenAI reported that no significant disruption reports were received for its other services, Codex and the application programming interface (API). This suggests that API-based services primarily used by enterprise customers and developers remained operational.
OpenAI is currently working to identify the root cause and has not yet disclosed a specific timeline for full restoration.