OpenAI Probes Performance Issues in ChatGPT and Codex

OpenAI is currently investigating significant performance issues impacting its popular AI services, ChatGPT and Codex. The company has acknowledged the problem, which has led to increased error rates across these platforms.

Ongoing Investigation into Service Problems

The issue was flagged as “Investigating” on OpenAI’s status page, indicating that while engineers have noticed irregularities, the exact cause and resolution timeline remain unclear. Both ChatGPT and Codex, crucial for AI-driven conversations and coding tasks, are affected.

The status update specifies the investigation is focused on these services, listing 15 components of ChatGPT and four of Codex currently affected. Users might face challenges such as failed requests, unexpected errors, or delayed responses.

Impact on Developers and Users

Developers relying on Codex for code generation and debugging may experience diminished reliability. OpenAI has yet to identify which specific features or regions are most impacted. The incident began at 08:28 PM on September 3, 2026, and details are still being determined.

OpenAI’s system-status dashboard continues to reflect overall service availability, but users should note that these figures do not confirm full functionality of all features. Partial outages can occur without affecting the entire service.

Industry Implications and Recommendations

This incident is part of a broader trend of AI service disruptions, similar to a recent outage at Anthropic’s Claude AI. Such issues highlight the risks for enterprises relying solely on external AI platforms for critical operations.

Organizations utilizing ChatGPT or Codex for essential functions should monitor OpenAI’s official updates. Activating alternative processes, like local tools and manual reviews, is advisable until normal service resumes.

Currently, OpenAI has not linked the performance problems to any cyberattack or infrastructure failure. The advisory merely confirms ongoing performance degradation and investigation efforts.

Users are encouraged to retry requests cautiously, avoid resubmitting sensitive data, and log error details for support inquiries.