"We are working on implementing a mitigation," OpenAI said as users around the world found ChatGPT inaccessible.
Scope and timing of the outage
The interruption began at approximately 8:00 PM ET on Wednesday, August 19, and quickly showed signs of global reach. Reports and checks indicated users in the United States and Europe were affected; the disruption prevented sign-ins, the creation of new accounts and the loading of existing chats. An update posted at 8:15 PM ET said the incident had been identified and remained ongoing roughly 14 minutes after it began.
User experience: what breaks when ChatGPT fails
The visible symptoms were consistent across reports. On chatgpt.com, sessions stalled at loading animations in the sidebar; attempts to send messages failed with "too many concurrent requests" errors. Users attempting to create new accounts or log in encountered failures during signup and authentication. The outage also prevented loading of previous conversations, leaving users unable to retrieve prior work or history hosted on the service.
This site is the portfolio.
OSINTSights runs on Cloudflare Workers, D1, R2, and Vectorize, with an AI pipeline on Hetzner ARM. Nubivance designed, built, and operates it. We do the same for clients.
See what we buildOpenAI API: multiple endpoints affected
The disruption was not limited to the web interface. OpenAI's status page listed as many as 12 API endpoints as having issues, indicating service degradation for programmatic access as well as interactive use. The simultaneous impact on both the public site and multiple API endpoints expanded the outage beyond individual browser sessions to any application depending on those API paths.
OpenAI's response and status updates
Public messaging from OpenAI acknowledged a login issue across the impacted services and described active work on implementing a mitigation. The 8:15 PM ET update reiterated that the incident was identified and ongoing, and added: "We continue to run into issues in our tests." Beyond that, the update conveyed that engineers were attempting fixes while monitoring the failure modes reported by users and automated checks.
How technologists, enterprises, and end users are affected
- Technologists and security teams: Developers relying on the OpenAI API saw multiple endpoints marked with issues, a concrete indicator that integrations and automated workflows may fail until those endpoints return to normal. The "too many concurrent requests" error reported on the web service is a clear symptom to surface in monitoring and incident dashboards.
- Affected enterprises and procurement leaders: Organizations using ChatGPT or the OpenAI API in production faced immediate service interruptions; the list of up to 12 impacted API endpoints on the status page gives procurement and SRE teams a discrete starting point for incident triage and vendor communication.
- End users and the general public: For individuals, the outage meant inability to sign in, create accounts, continue conversations or retrieve prior chats. The failure to load previous conversations directly interrupts the continuity of work and personal use stored on the service.
The record in the status updates is straightforward: the outage began at about 8:00 PM ET on August 19, affected both the web interface and multiple API endpoints, and remained unresolved as of an 8:15 PM ET update while OpenAI worked on mitigations. The concrete facts — stalled sidebar loads, "too many concurrent requests" errors, login and signup failures, and as many as 12 API endpoints listed as impacted — define the operational contours of the incident and the immediate tasks facing engineers and affected customers.
Read the original report: https://www.bleepingcomputer.com/news/artificial-intelligence/openai-confirms-chatgpt-is-down-as-logins-and-signups-fail/