OpenAI confirmed at around 7am this morning that the site of its popular AI chatbot, ChatGPT, was experiencing an outage, following reports of elevated error rates.
It said at the same time that the root cause has been identified, which it later revealed to be “elevated database CPU usage impacting the site over the last several hours.”
The issue seems to have started on Tuesday morning as the company reported elevated error rates, lasting for about an hour or so, across all its services.
The incident resurfaced on Wednesday afternoon, affecting text-davinci-003, the moderation endpoint, gpt4, ChatGPT, whisper and turbo engine. Incident reports on Downdetector also surged, and did so again this morning when today’s outage was detected.
OpenAI said an hour later that the issue has been resolved and that it continues to monitor the ChatGPT site to ensure the incident does not recur.
Last Thursday, the Microsoft-backed company experienced a similar issue that lasted about three hours, impacting the ChatGPT site and API customers. All systems were operational after the company said it “implemented fixes to prevent similar issues in the future.”
This story will be updated as more information becomes available.