OpenAI experienced a significant service outage on September 3, 2026, affecting both ChatGPT and its Codex coding assistant across multiple platforms and interfaces. The company opened an incident report after detecting elevated error rates across 19 components, with the investigation beginning at 2:56 PM UTC. The outage touched every major surface where developers and users access these tools, from web interfaces to command-line tools and editor extensions.

Which Codex and ChatGPT Components Were Affected?

The outage was remarkably broad in scope. On the Codex side, all four major access points experienced degraded performance: the web interface, the application programming interface (API) that developers use to build with Codex, the command-line interface (CLI) for terminal-based coding, and the Visual Studio Code extension that integrates Codex directly into the popular code editor.

The ChatGPT side of the incident was even more extensive, affecting 15 different components. These included core functionality like conversations and login systems, newer features such as voice mode and deep research capabilities, and workspace-focused tools like ChatGPT Work and the Agent feature. The outage also impacted file uploads, image generation, and the ChatGPT Atlas browser tool.

• Codex Components Down: Web interface, API, command-line interface, and VS Code extension all reported degraded performance
• ChatGPT Core Services: Login, conversations, voice mode, and file uploads experienced elevated error rates
• Advanced Features Affected: Deep research, the Agent feature, image generation, and ChatGPT Work all went offline
• Workspace Tools Impacted: Sites, connectors, and apps integrations stopped functioning normally

What Stayed Online During the Outage?

Notably, OpenAI's broader API services remained unaffected. The company's APIs group, which includes chat completions, embeddings, fine-tuning, audio processing, and the Sora video generation tool, showed no degradation. The FedRAMP-compliant services and the Ads Platform also continued operating normally.

This distinction is important for developers who rely on OpenAI's underlying API infrastructure. While users trying to access ChatGPT directly or developers using Codex through its various interfaces faced problems, those building applications through OpenAI's core API endpoints could continue their work uninterrupted.

How Serious Was the Outage?

OpenAI classified the incident as "degraded performance" rather than a complete outage, meaning services remained partially accessible but with elevated error rates. The company's status page showed that ChatGPT had maintained 99.62 percent uptime over the June through September 2026 period, while Codex showed 99.98 percent uptime over the same window.

However, these aggregate figures mask the real-time impact. When OpenAI published its initial update at 2:58 PM UTC, the company stated only that it was "investigating the issue for the listed services," providing no timeline for resolution or explanation of the root cause. The incident represented the second service disruption OpenAI logged for ChatGPT on that single day, following an earlier "ChatGPT Work Mode High Error Rates" incident that was resolved at 12:10 AM.

How to Stay Informed About OpenAI Service Issues

• Monitor the Status Page: Visit OpenAI's official status page regularly to check real-time incident reports and component health for ChatGPT, Codex, APIs, and other services
• Subscribe to Updates: Use the subscription option on OpenAI's status page to receive notifications when new incidents are reported or when existing incidents are updated
• Check Specific Components: If you use a particular tool like the Codex CLI or VS Code extension, verify that specific component's status rather than assuming all services are down
• Review Incident History: Examine OpenAI's incident history index to understand patterns of service disruptions and which components are most frequently affected

The September 3 incident underscores a broader reality for developers and teams relying on cloud-based AI tools: even widely-used platforms experience service degradation. OpenAI's incident reporting shows that the company had already dealt with two separate ChatGPT issues on that single day, plus an account creation problem the day before. For teams using Codex as part of their development workflow, having a backup plan or understanding which components are most critical to your work becomes essential.

As of the latest update in the incident record, OpenAI had not identified a root cause or provided a timeline for full resolution. The company's status page remained the designated channel for further updates, and users were encouraged to monitor it for new information as the investigation continued.