ChatGPT users around the world ran into a wall of 404 errors, with the chatbot, its website, and OpenAI’s coding tool Codex all throwing “Not Found” messages instead of responding to prompts, even as OpenAI’s own status page insisted everything was running fine.
Screenshots shared widely on X showed some users hitting the same cryptic error: “unexpected status 404 Not Found: Unknown error,” pointing to OpenAI’s backend API. One user, posting to X as @ollicook, tagged OpenAI directly and noted the mismatch between what people were experiencing and what OpenAI’s status page was reporting. The status page showed a reassuring green checkmark and “We’re fully operational,” listing 99%+ uptime across APIs, ChatGPT, Codex, and other services, even as the error reports piled up.
For some, the problem went beyond a backend error message. Visiting chatgpt.com directly returned a plain 404 page reading “This chatgpt.com page can’t be found,” the kind of error typically seen when a domain or route doesn’t exist at all, rather than a server being merely overloaded.
Codex, OpenAI’s AI coding assistant, wasn’t spared either. A user posting as @JohnMuyno, in a thread translated from Chinese, said Codex threw the same “404 Not Found: Unknown error” message when hitting its responses endpoint, and joked that with ChatGPT, Codex, and other tools all down at once, it felt like users were being pushed toward Google’s Gemini instead.
OpenAI has not put out a public statement specifically addressing this round of 404 errors. As with previous incidents, the company’s typical pattern has been to acknowledge elevated errors on its status page once reports cross a certain threshold, then post updates as a mitigation is identified and rolled out, with most disruptions resolved within a few hours.
Given how central ChatGPT and Codex have become to daily workflows for millions of users, even a short-lived error like this one is a reminder of how quickly things grind to a halt when OpenAI’s backend stumbles, and how big a gap can open up between what OpenAI’s dashboards report and what users are actually seeing on their screens.