OpenAI outage took down ChatGPT, API and Codex together

ChatGPT, OpenAI's API and Codex all failed together on the morning of July 25, per BleepingComputer. That matters for teams whose fallback plan is the API when ChatGPT breaks: it went down in the same outage.

Downdetector reports peaked around 1,535, about 80% against ChatGPT and 8% each against the mobile app and the API. Saved conversation history disappeared for many accounts before service came back, with recovery reaching the monitoring stage in about an hour.

The 503 errors carried an internal label, biscuit_baker_service_me_circuit_open, that BleepingComputer's report didn't explain. A name like that points to a circuit breaker tripping upstream of chat, API and Codex traffic alike. The report doesn't say which shared service tripped it.

That's the real story: three products that look like separate surfaces, a chat app, an API, a coding tool, went down on the same clock. That only happens if they sit behind one shared dependency, whether that's auth, a request router, or the circuit breaker itself.

If ChatGPT, the API and Codex share one circuit breaker, the API isn't a real fallback for a ChatGPT outage. It just failed alongside it.