OpenAI spent much of Monday investigating a sweeping outage that knocked ChatGPT, its Codex coding tool and its API Platform offline or into a degraded state for users across the United States, the United Kingdom and India. Reports on Downdetector first spiked at around 10:05 AM ET (3:05 PM GMT), and OpenAI's status page quickly moved to a 'partial outage' state citing 'degraded performance' across multiple products.
What broke
The symptoms were inconsistent but broad. Some users could not sign in at all, while others reached the chat interface only to see messages fail to send or prior conversations refuse to load. Voice mode and image generation went dark for a subset of paying customers, and developers hitting the API reported elevated error rates. Codex, the coding product that sits on top of the same backbone, was listed alongside ChatGPT and the API on OpenAI's incident page — underscoring how tightly coupled the company's public surface has become to a single shared infrastructure.
Scale of the disruption
Downdetector traffic suggests the incident landed hardest on the UK, where more than 7,600 reports piled up around peak. US users logged roughly 1,700 reports in the same window, while Indian users registered around 756 reports. For a service that OpenAI has publicly said handles hundreds of millions of weekly users, the visible complaints are almost certainly a small fraction of the real blast radius, particularly for enterprises whose production workflows now route through the API.
OpenAI's response
Public communication was sparse. The status page updates confined themselves to acknowledging the problem and saying engineers were 'continuing to investigate.' No executive statement, no postmortem, and no root cause had been published by the time service began recovering. Tom's Guide reported that ChatGPT was broadly usable again by around noon ET, roughly two hours after the initial spike, though some users continued to hit intermittent errors into the afternoon.
Why this one stings
Outages are routine for any cloud product, but the shape of this one is what matters. OpenAI has spent the past year positioning ChatGPT and its API as critical infrastructure — embedded in enterprise workflows, coding agents like Codex, and a growing set of third-party products. Every minute those surfaces are down now cascades into missed customer tickets, stalled CI jobs and frozen agentic workflows downstream.
It also lands at an awkward moment competitively. Rivals including Anthropic and Google have been aggressively pitching enterprise buyers on reliability and model diversification, and a visible multi-hour wobble across ChatGPT, Codex and the API gives procurement teams fresh ammunition to argue for second-vendor strategies. OpenAI's silence on the cause is unlikely to help; customers running production traffic through the API will want a detailed postmortem, not just a green checkmark.
