OpenAI reported a widespread service disruption on April 20, 2026, affecting its flagship ChatGPT platform and the Codex developer tool. The outage began at approximately 10:05 a.m. ET, resulting in a total loss of access for many users across global regions, including North America, Europe, and Asia. According to the official OpenAI status dashboard, the company began investigating reports of users being unable to load services shortly after the initial spike in connectivity failures was detected.
Technical data from outage-tracking service Downdetector indicated a sharp increase in incident reports, peaking at over 8,000 in the United Kingdom and approximately 2,373 in the United States. Users in India also reported significant disruptions, with over 900 incidents logged by late evening local time. The issues were not limited to the web interface; approximately 11 percent of reported problems originated from the mobile application, while 79 percent of users flagged general service outages. A smaller subset of users, roughly 5 percent, reported specific failures related to web browser compatibility and pinned chat synchronization.
The disruption impacted a broad range of features, including user login, search functionality, and the platform’s integrated voice mode. Users attempting to utilize the voice conversation feature reported that the interface would abruptly stop or display an error message stating it could not load conversations. Codex, which powers many third-party developer integrations, was similarly unresponsive, disrupting automated coding workflows. OpenAI confirmed that the outage affected all tiers of service, including the free version, ChatGPT Plus, and institutional workspaces such as ChatGPT Enterprise and ChatGPT Edu.
During the peak of the incident, OpenAI’s system uptime for ChatGPT fell to 99.85 percent, a notable dip for a service that typically maintains higher availability. In a technical update, the company noted a specific secondary issue impacting ChatGPT Business accounts, where users might encounter errors for up to an hour following account upgrades or the addition of new seats. By late afternoon, OpenAI stated it had applied a mitigation strategy and was monitoring the recovery process to ensure stability across its global server network.
While the primary services began to stabilize, engineers continued to investigate the root cause of the connection failures. This event follows a series of intermittent degraded performance warnings issued by the company throughout March and early April 2026, including a significant warning on April 2 that lasted over three hours. OpenAI has not yet released a full post-mortem report regarding the specific infrastructure or software failure that led to the April 20 disruption, though the company’s status page currently lists the incident as under active monitoring.