OpenAI Restores Services After September 3 Outage Affecting ChatGPT and Codex

OpenAI's systems are back to full operational status after a service degradation event on September 3 that affected key functionalities across ChatGPT and Codex. The disruption, which lasted for nearly four hours, impacted at least 19 components including conversations, image generation, and login capabilities.

The incident began around 10:58 AM ET and was resolved by 2:56 PM ET, leading to a spike in error rates that locked users out of essential features. Earlier that day, OpenAI had already reported a separate issue with ChatGPT's Work Mode, indicating a challenging day for the platform.

OpenAI confirmed the resolution of the issue through updates on its public status page, which reflected uptime figures of 99.63% for ChatGPT and 100% for Codex over recent periods. This incident is part of a growing pattern of service disruptions, with hundreds logged since early 2025, raising concerns about the reliability of OpenAI's infrastructure.

Despite the timing coinciding with promotional activities for the upcoming GPT-6 Astra model, OpenAI has not indicated a direct correlation between the outage and its launch preparations. The company has focused on operational transparency and effective post-incident management, successfully restoring services to their full operational status.

New Developments on AI Outages

As of September 3, 2026, multiple AI models including Claude, Grok, ChatGPT, and Gemini are experiencing significant outages. Users have reported that fifteen different services from OpenAI are currently facing issues, which the company is actively working to resolve.

Claude's status tracker has indicated that its Opus 4.8 and Opus 5 models are among those affected, while others have reportedly returned to a baseline error rate. In contrast, the Gemini 3.8 model appears to still be operational according to some users.

X has confirmed that Grok is experiencing service degradation, and Cursor has reported that all Grok models, automations, and related services are impacted. Despite claims from Google's status page that Gemini is functioning normally, Down Detector has received numerous reports indicating that some users are unable to access it.

As of now, none of the affected AI models have publicly addressed the outages on social media platforms.

New Insights on the September 3 AI Outage

• ChatGPT outage reports surged from 5,000 to over 22,000 within 10 minutes on the morning of September 3, 2026.
• Users began reporting issues with Grok and Claude around 9 a.m. Eastern Time, with ChatGPT complaints peaking 90 minutes later.
• Charles Hoskinson, founder of Cardano, suggested that the simultaneous outages of Claude, ChatGPT, and Grok appeared to be the result of a coordinated attack by a nation-state.
• OpenAI confirmed elevated error rates across ChatGPT and Codex, while Anthropic's Claude faced capacity issues.
• Reports indicated that Grok and Claude were both using significant portions of the same compute resources, raising questions about potential shared vulnerabilities.
• The incident has sparked discussions about the need for decentralized AI systems to mitigate risks associated with concentrated compute resources.
• Anthropic's compute facility in Memphis, which houses a large number of Nvidia chips, was highlighted as a potential source of the outage, although no direct link has been established.