ChatGPT desktop outage locks out Work threads for over six hours

On October 7, ChatGPT desktop users suddenly lost the ability to start new Work threads. The problem dragged on for more than six hours after a faulty app update broke one of the platform's main features. OpenAI called it a major incident. The breakdown also exposed a clear gap in how the company supports different operating systems: Windows users had to wait two hours and 24 minutes longer than those on macOS and Linux before they could get back to work.

The trouble started at 05:31:56 UTC. OpenAI's status page pointed to version 26.1002.51308 as the source of the problem. Within 22 minutes, OpenAI admitted the issue and began rolling back to the previous version. But the real split came later. Linux and macOS users got a fixed app at 09:28:55 UTC. Windows users, meanwhile, were told only that a fix was in progress. The Windows update finally arrived at 11:53:35 UTC. That closed the incident after six hours, 21 minutes, and 39 seconds of downtime for those affected.

Platform gap and OpenAI's response

OpenAI left Windows users hanging with no reason for the extra delay. The company never said how many accounts were hit or where the failures were worst. The app itself stayed online; OpenAI just rolled back to an older version and then pushed out a patched update. The only advice for users was to update to the latest version to get Work threads working again. According to an official OpenAI incident report, the company marked the issue as resolved after shipping corrected builds for Linux, macOS, and Windows, and told everyone to update their app.

Work mode is one of two main features in the ChatGPT desktop client. It lets users run projects side by side, manage files, and keep workflows going. When new Work threads stopped working, a big part of the app's purpose was lost-especially for people who depend on it for ongoing projects or teamwork. OpenAI confirmed the bug was tied to version 26.1002.51308. Fixes rolled out in stages: first to Linux and macOS, then to Windows hours later, as shown in the OpenAI status update.

Incident context and severity

This was not the only technical problem OpenAI logged that day. October 7 saw four other incidents, but only the Work thread failure got a "major" label. Other issues-workspace agent slowdowns, Codex thread creation failures, and a wider Codex and Work mode slowdown-were all rated minor and fixed in under two hours. The last major incident before this happened on October 1 and lasted just 53 minutes, much shorter than the Work thread outage.

OpenAI doesn't just look at how long an outage lasts. The desktop build was called major because starting new Work threads was totally broken, not just slowed down. For comparison, a five-hour, 21-minute outage on September 29 that hit ChatGPT, Codex, and the API was rated minor. That shows OpenAI puts total feature loss above broad but partial slowdowns when deciding severity.

What it meant for users

For people who rely on ChatGPT's Work mode every day, the outage was more than a hassle. It brought work to a standstill. OpenAI gave no details about how many were affected or where, and Windows users had to wait even longer for relief. This isn't the first time Work mode has stumbled. As reported earlier, earlier errors have already disrupted scheduled tasks for those who count on this feature.

OpenAI's only fix was to tell users to update their app. There was no extra explanation or compensation. The outage shows how even advanced AI tools can grind to a halt because of a single bad update-and not everyone gets help at the same speed. Until OpenAI brings more consistent and open incident handling across platforms, users-especially on Windows-face the risk that key features can disappear without warning, and recovery may not come equally fast for all.