OpenAI Outage History

OpenAI began investigating increasing error rates in Codex on 3:56 PM UTC, marking a degraded performance for Codex-related components. The impact affected ChatGPT / Codex, ChatGPT / Deep Research, and ChatGPT / Login, with users experiencing login issues, errors when using the GitHub connector with Deep Research, and general Codex-related failures. OpenAI issued status updates as engineers worked to implement a fix, and services gradually returned to normal later that day. The outage lasted about 4 hours and 41 minutes, resolved by 8:37 PM UTC. User reports were present but limited, indicating a localized impact rather than a service-wide outage.

OpenAI's Codex feature on the web was unavailable for all users, impacting web-based code generation accessed through the ChatGPT interface. The outage began at May 27, 2025 9:41 PM UTC and was resolved by May 27, 2025 10:52 PM UTC, lasting about 1 hour and 11 minutes. The issue affected the ChatGPT / Codex component and disrupted workflows relying on Codex in the web interface for users worldwide. OpenAI documented the incident on its status page as Codex unavailable and issued updates as engineers investigated and restored service. The provider released two official status updates during the incident, and service was restored after the root cause was addressed. The impact was global, affecting users attempting to access Codex on the web.

OpenAI experienced a subscription access outage that prevented some ChatGPT Pro and Team customers from accessing their paid features due to a billing-system issue. The problem appeared across a broad set of ChatGPT components, blocking access to subscriptions and related functionality for impacted users. The incident began on May 26, 2025 6:47 PM UTC and was resolved by May 27, 2025 5:28 PM UTC, lasting roughly 22 hours and 41 minutes. Users reported difficulties with accessing or renewing subscriptions, signing in, or using paid features.

The OpenAI Codex service experienced degraded performance affecting task creation and PR creation starting at May 26, 2025 8:11 AM UTC. Observed elevated error rates were potentially tied to a degradation in the GitHub API, impacting users attempting to create Codex tasks or PRs via ChatGPT / Codex workflows. The outage lasted approximately 1 hour 35 minutes, ending at May 26, 2025 9:46 AM UTC. The incident involved the ChatGPT / Codex component and prompted 2 status updates as OpenAI investigated and resolved the issue.

The provider communicated through status updates, noting the degraded operation and subsequent recovery as the GitHub API issues were addressed. Reliability returned to baseline after the final update.

The outage began at May 23, 2025 6:36 PM UTC and was resolved by May 23, 2025 6:42 PM UTC, lasting about 6 minutes. Users attempting to use OpenAI Codex features within ChatGPT and related APIs experienced elevated errors or degraded responses during the incident. The issue affected Codex functionality but did not indicate a broader disruption to all OpenAI services.

OpenAI issued 2 status updates during the incident, confirming mitigation steps were underway and that Codex functionality returned to normal shortly after. The incident was short, with operations fully restored shortly after the brief outage.

The outage began on May 22, 2025 9:41 AM UTC and was resolved by May 22, 2025 12:27 PM UTC. OpenAI reported elevated error rates across ChatGPT Web and mobile clients, with users experiencing API errors and incomplete responses. The issues affected ChatGPT / Web, ChatGPT / Mobile - iOS, and ChatGPT / Mobile - Android, resulting in degraded performance and intermittent access for a portion of users. The outage lasted approximately 2 hours and 46 minutes.

OpenAI published two status updates during the incident, first noting an investigation and later confirming restoration of normal service as responses returned to typical operation. User reports were modest but aligned with the observed errors, including 503 Server Unavailable messages.

OpenAI reported high error rates for Codex tasks, causing degraded performance for users executing Codex-related work in ChatGPT/Codex. The outage began at 1:16 AM UTC and was resolved by 1:36 AM UTC, lasting about 20 minutes. OpenAI issued two status updates during the incident, noting investigations and a subsequent recovery that returned Codex functionality to normal.

OpenAI reported a brief degraded state affecting the ChatGPT login flow, with users experiencing login problems when signing in.

The incident began on 2:41 AM UTC and was resolved by 3:51 AM UTC, lasting about 1 hour and 10 minutes. The OpenAI status page showed two updates during the incident, noting that login issues had been identified and mitigations were being implemented.

No other OpenAI services were reported affected, and sign-in functionality returned to normal after the mitigation was applied.

OpenAI reported a brief degradation in ChatGPT Voice mode starting at 10:01 AM UTC, with users seeing increased errors in voice transcription. The issue affected voice input for ChatGPT users and was resolved by 10:12 AM UTC after about 11 minutes. OpenAI issued two status updates during the incident, confirming the problem and its restoration. User activity began around 9:30 AM UTC with about one report, indicating limited but noticeable impact on users relying on voice transcription. The outage appeared global and did not indicate a full platform outage, but rather a short lived degradation in ChatGPT Voice mode.

OpenAI reported a brief degradation in ChatGPT Search where users experienced elevated failures and high latency. The outage began at 1:26 PM UTC and was resolved by 1:42 PM UTC, lasting 16 minutes. The disruption affected the ChatGPT / Search component, causing slower search results for a subset of users.

OpenAI issued two status updates during the incident, noting investigation and subsequent restoration.