ChatGPT, API and Codex Go Down Together for the Fourth Time in Four Days

What happened

On the morning of Saturday, July 25, 2026, ChatGPT, OpenAI's APIs and Codex simultaneously reported elevated error rates and outages starting at around 5 a.m. ET. The disruption hit users worldwide, including in the United States, India and Europe, according to BleepingComputer and TheNextWeb. OpenAI publicly confirmed that ChatGPT was down.

What users experienced

Users reported the ChatGPT sidebar getting stuck on loading animations and being unable to send messages, with the system returning a too many concurrent requests error. According to Business Upturn and Sunday Guardian Live, which tracked reports through Downdetector, some users specifically saw 503 errors carrying the internal label biscuit_baker_service_me_circuit_open, which kept requests from ever reaching OpenAI's servers.

The fourth outage in four straight days

Saturday's incident was not isolated. TheNextWeb reported it as OpenAI's fourth service disruption in four consecutive days, a pattern that points to structural instability in the company's infrastructure rather than a single event. The short interval between failures matters because ChatGPT, its API and Codex underpin third party commercial products at global scale.

Authentication suspected as the root layer

Multiple reports pointed to OpenAI's authentication service as the likely root cause layer, since a failure there would explain why chat, API and Codex went down at the same time. OpenAI had not publicly detailed a confirmed root cause as of publication.

OpenAI's response

The company's status page moved from investigating to monitoring within about an hour, indicating a mitigation had been applied while the team watched for full recovery. Still, the fact that this was the fourth occurrence in four days suggests the underlying issue was not fully resolved between events.

Why this matters for Brazilian agencies and SMBs

For Brazilian automation agencies and small businesses that have built chatbots, sales agents and support flows on top of the ChatGPT API or Codex, four outages in four consecutive days are a concrete, dated reminder that even the largest AI vendor is not immune to repeated infrastructure failures. Any production system serving end customers, especially live customer service, where a failure is visible to the client's own customers in real time, needs a documented fallback plan: a secondary model or provider, a graceful degradation message and a human escalation path. This is not theoretical, it is practical risk management, and OpenAI's recent track record makes the lesson unavoidable.