ChatGPT back online after facing major outage: Here's what happened

OpenAI's ChatGPT has been restored after a major global outage disrupted access for thousands of users, with error reports peaking overnight before services were gradually stabilised

OpenAI’s ChatGPT is back online after a major outage that reportedly impacted thousands of users across the globe. According to a real-time problem and outage monitoring platform, DownDetector, the majority of outage incidents reported in India were logged at around 01:55 AM (IST) on February 4. Out of the reported cases, 89 per cent were related to ChatGPT itself. 

What went down

OpenAI faced a major global outage on February 4. According to media reports, there were over 12,000 reports from the US alone on the outage. In India, the peak number of ChatGPT outages reported were logged at 01:55 AM IST, with the number of reports standing at 1,516.

 

OpenAI admitted this on ChatGPT’s status page at around 2:00 AM (IST), where it wrote, “Elevated error rates for ChatGPT and Platform,” which is a developer-focused portal for accessing OpenAI's models. The company added that they are working to fix the issue. After around 25 minutes of posting this, OpenAI shared another update in which it wrote, “Elevated error rates for finetuning jobs,” without explaining what these jobs were. 

 

At around 3:00 AM IST, OpenAI reportedly fixed the issue, and thereafter, the platforms gradually began working properly.

What might have been the cause of this outage?

OpenAI rolled out its Codex app for macOS on February 2, and CEO Sam Altman said the release saw strong traction, with more than 200,000 downloads on its first day. A day later, Apple pushed an update to Xcode, introducing agent-based coding support for Codex and Claude.

 

9to5Mac noted that although there is no indication that these launches directly triggered the ongoing ChatGPT outage, the surge in Codex usage may have added unexpected strain on OpenAI’s systems. We won’t be able to know the exact reason until OpenAI officially releases a statement talking about the issue.