OpenAI’s ChatGPT experienced unstable access and an increase in errors on September 3 (local time).
Other major AI services, including Claude and Grok, also experienced outages around a similar time.
The fact that multiple services stopped at the same time and the conclusion that they had a single cause must be considered separately.
3-line summary
1. Major AI services, including ChatGPT, experienced simultaneous outages
2. ChatGPT errors increased at 11 a.m. U.S. Eastern Time
3. A common cause has not been confirmed
Increased Errors and Functional Impact on ChatGPT and Codex
OpenAI said it was investigating an increase in errors on ChatGPT and Codex. According to AI Times, the outage was not limited to the chat function. ChatGPT’s major features overall were affected, including login, file uploads, voice mode, search, deep research, and image generation.
The Electronic Times reported that ChatGPT access became unstable at around 10:20 a.m. on September 3, U.S. Eastern Time, and that 3만7000 reports were submitted to the outage-tracking site Downdetector at 11 a.m. The disruption was not simply a matter of response generation becoming slower; inconvenience continued from the login stage required to start the service through the process of using various features.
OpenAI later announced via its status page that the issue involving increased errors on ChatGPT and Codex had been resolved. Users of Codex’s remote-control feature were also told that they might need to reconnect their remote-control devices. In other words, even after the service was restored, some users had to check the connection status of individual features again.
A Day When Claude and Grok Also Stopped: An Unconfirmed Link
Around the same time, Anthropic’s Claude and SpaceX’s Grok also experienced access outages. Claude explained that the issue was an infrastructure problem that caused a partial interruption across the service and said it had resolved the problem at around 12:15 p.m. U.S. Eastern Time. Grok was also reported to have returned to normal afterward.
Yonhap News TV reported that these services had generally regained stability after 12 p.m. U.S. Eastern Time. It is unusual that services operated by different companies stopped at a similar time, but this does not in itself mean that the outages originated from the same cause.
The possibility of a major cloud infrastructure problem and issues related to Microsoft Azure were also raised. However, AI Times reported that it had not been confirmed whether the outages affecting the three services were actually connected to a single cause. The conclusion confirmed in this incident extends only to the fact that major AI services experienced outages around a similar time and were restored one after another.
References
Tags #ChatGPT #ChatGPTOutage #ChatGPTErrors #OpenAI #Codex #Claude #Grok #AIOutage #ServiceOutage #Downdetector #GenerativeAI #AIService