ChatGPT goes dark worldwide as OpenAI scrambles to fix login failures
A global outage starting late Wednesday locked users out of ChatGPT, Codex and a dozen OpenAI API endpoints.
Key points
- ChatGPT began failing worldwide at roughly 8 PM ET on Wednesday, 19 August, blocking logins, signups and chat loading.
- OpenAI's coding assistant Codex and about 12 endpoints on the OpenAI API were hit in the same incident.
- OpenAI acknowledged the problem on its status page within about 14 minutes and said it was working on a fix.
- Affected users saw endless loading spinners and "too many concurrent requests" errors when trying to send messages.
- This is a service outage, not a security breach: no data theft has been reported.
ChatGPT stopped working for people around the world on Wednesday night, and OpenAI has confirmed the problem.
The trouble started at about 8 PM Eastern Time on 19 August. Users in the United States, Europe and beyond found they could not sign in, could not create new accounts, and could not open any of their old conversations.
When people tried to load the site, the sidebar spun forever. Sending a message threw back a "too many concurrent requests" error, which is the system's way of saying it is swamped and cannot take any more work.
The outage was first reported by BleepingComputer.
What actually broke?
OpenAI's login system appears to be the main casualty, and that failure cascades into almost everything else the company offers. Without a working sign-in, nobody can reach their chats, and developers cannot call the tools that sit behind ChatGPT.
On its public status page, OpenAI said it had "identified" the issue and was "working on implementing a mitigation." That is company shorthand for: we know what is wrong and we are trying a fix.
The knock-on damage is wider than the chatbot itself. Codex, OpenAI's coding assistant that helps programmers write software, went down at the same time. So did around 12 endpoints on the OpenAI API, the plumbing that other companies use to build ChatGPT-style features into their own apps.
| Detail | What we know |
|---|---|
| Start time | ~8 PM ET, Wednesday 19 August |
| Acknowledged by OpenAI | ~8:15 PM ET (about 14 minutes in) |
| Services affected | ChatGPT, Codex, OpenAI API (~12 endpoints) |
| Regions affected | Global, including US and Europe |
| Cause | Login and authentication failures |
Is this a hack?
No. Nothing in OpenAI's status updates or the early reporting points to an attack. This looks like a classic availability failure, most likely inside the authentication service that checks who you are when you log in. Outages of this shape happen to every large cloud service eventually, from email providers to banks.
That matters because a lot of people, understandably, assume the worst when a login page stops working. There is no indication that passwords, chat histories or payment details have been exposed.
What should users do?
Wait it out. There is nothing a normal user can do from their side to make ChatGPT load faster. Refreshing the page repeatedly or trying to reset a password will not help, and hammering the login button is part of what generates those "too many concurrent requests" errors in the first place.
A few sensible habits for anyone who relies on ChatGPT for work:
- Keep a local copy of anything important you have generated. Web services go down, and your chat history is not a backup.
- Be wary of any email claiming to be from OpenAI during an outage. Criminals often push fake "account recovery" messages when a big service is down, hoping panicked users will hand over their passwords.
For developers whose products depend on the OpenAI API, this outage is a reminder that a single upstream provider is a single point of failure. Building in a fallback, even a graceful error message, tends to age well.
OpenAI had not published a root cause at the time of writing. We will update this story as the company shares more.
