OpenAI spent the morning investigating elevated latency that disrupted ChatGPT and its API, leaving thousands of users unable to load chats or get responses. By 8:15 am PT the company’s system status page showed the incident as active.
The immediate scale was plain: a supplementary report said ChatGPT was down for thousands of users, and Downdetector logged over 1,000 reports at the time of writing. OpenAI acknowledged the outage on its server page and said users were experiencing elevated errors for the impacted services, and that team members were investigating elevated latencies across the board for API and ChatGPT.
For many users the interruption looked like blank chat windows or endlessly loading screens. The supplementary article said users in India had particular trouble accessing ChatGPT and some of the service’s coding features, with failures to load chats or generate responses that matched the error profile OpenAI described on its status page.
The timeline the company posted tracked the incident in near real time. At 8:15 am PT the status page showed OpenAI was investigating elevated latencies across API and ChatGPT. By 9:15 am PT the ChatGPT issue was reportedly resolved, but OpenAI noted the API slowdown was still occurring. A separate FedRAMP issue that had been monitored received a fix and was removed from the status page. At 10:31 am PT the issue appeared to be resolved.
OpenAI said it was working on implementing a mitigation while the disruption continued. The company removed the FedRAMP issue from the status page after the fix was applied, and by mid‑morning its updates indicated the immediate errors had been cleared for the services it had flagged.
The friction in this episode is clear: ChatGPT itself was marked as fixed well before all effects on the API had cleared. That split matters because developers and services that depend on the API can continue to see degraded performance even after conversational access looks restored to end users. OpenAI’s acknowledgement of elevated errors and its note that it was deploying a mitigation underscore that the recovery was staged rather than instantaneous.
This outage also illustrated how visible these interruptions have become. Thousands of affected users, reported problems on outage trackers, and descriptions of blank or endlessly loading chats gave a tangible shape to what the status page called elevated latency. Even with the FedRAMP fix applied and the separate entry removed from the status page, the staggered timeline means some customers experienced a longer window of disruption than the initial alert suggested.
By 10:31 am PT the incident appeared to be resolved across the board, but the core question now is operational: will the mitigation that OpenAI said it was implementing prevent a recurrence of elevated latencies for the API customers who continued to see slowdowns after ChatGPT access was reported fixed? The company’s timeline shows the problem was contained this time, but the staggered recovery highlights the dependency risk for services built on the platform—and whether that risk is permanently reduced will be answered by the next hours and days of traffic and monitoring.
