• A major outage at OpenAI on November 18 rendered ChatGPT and key platform services inaccessible for thousands of users globally.
  • The disruption, lasting approximately five hours, was primarily caused by failures in API batch processing and file upload functionalities.
  • The incident underscores the operational risks for businesses that have become deeply embedded with a single provider's AI infrastructure.

OpenAI’s services, including the widely used ChatGPT and its developer platform, experienced a significant outage on November 18, 2025, disrupting businesses and users for about five hours. The company’s status page indicated that the issues began at approximately 3:53 PM, with engineers identifying problems specifically related to Batch API jobs getting stuck during finalization and file uploads consistently failing.

The failure of these core services had an immediate impact. Businesses relying on automated batch processing for their AI workloads found operations stalled, while regular users were unable to attach documents or images to their ChatGPT queries. The outage triggered a surge of user reports on third-party monitoring sites and social media platforms, highlighting the broad dependency on OpenAI's ecosystem. A person familiar with the company's internal response said engineers had to manually intervene to clear the blockages in the batch job system.

Service was gradually restored throughout the evening, with file upload functionality coming back online by 8:55 PM. By later that night, most impacted services had recovered, according to the company’s updates. OpenAI did not immediately respond to a request for further comment on the root cause of the failure.

This is not the first time OpenAI has faced service instability, but the duration and specific nature of this outage have amplified concerns about the resilience of cloud-based AI infrastructure as it becomes more critical to daily business functions. The incident serves as a stark reminder of the fragility of complex, large-scale systems under load. In the short term, experts suggest that enterprises heavily invested in a single AI provider may need to reassess their contingency plans and push for more transparent incident reporting and stronger service-level agreements.

Correction: An earlier version of this article misstated the time for full service restoration. While core functionality was restored by 8:55 PM, the company confirmed most services were stable later that evening.