ChatGPT Work Mode Hit by New Errors, Disrupting Scheduled Tasks
On October 5, 2026, OpenAI flagged a new problem in ChatGPT Work Mode. Users faced errors that could break scheduled tasks without any clear sign. This time, the trouble went beyond real-time prompts. Automated jobs set to run on a schedule might have failed quietly. Many users had no idea unless they checked the results themselves.
OpenAI's help documentation specifies that scheduled tasks in ChatGPT Work Mode can be triggered by webhooks from Gmail, Slack, or GitHub activity, enabling integration with external workflows.
Silent failures put users at risk
The real risk is simple. These failures are invisible. Scheduled tasks in ChatGPT Work can run once, repeat, or trigger on changes. Many power key workflows. When they fail, there's no warning. OpenAI tells users to "Go to Scheduled to review results and schedules." No error message does not mean success. For anyone who relies on time-sensitive automation, manual checks are now a must.
OpenAI's advice is direct. If a task should have produced something important, check its last run. Rerun it if needed. If a task is paused or waiting for approval, review the Scheduled section. The risk is real. Missed or broken outputs can slip by until it's too late.
Instability keeps coming
This isn't the first time. ChatGPT Work Mode has seen repeated trouble since September 2026. There were high error rates on September 10. A 1% failure rate hit on September 11 and 12. More issues followed on September 14, affecting both Codex and Work. Errors came back on September 15, 16, and 17. On September 22, a bigger outage hit several subscription tiers. Another major incident struck on September 29, disrupting ChatGPT, Codex, and the Agents API. After that, OpenAI promised a full root-cause report within five business days. But the problems have kept coming into October.
OpenAI's help documentation details that task limits for scheduled jobs vary by plan: Free and Go users can have up to 3 active tasks, Plus users up to 5, Business and Edu up to 10, and Pro and Enterprise up to 15 active tasks.
What users need to do now
OpenAI says the October 5 issue is under control. Still, users must check that their scheduled jobs actually ran. Anyone who depends on ChatGPT Work for automated, time-sensitive work should review their task logs now. Rerun any jobs that failed or gave incomplete results. The risk is clear. A failed chat is obvious. A failed scheduled task can go unnoticed. Entire workflows could break.
These repeated Work Mode failures are not just technical glitches. They show a deeper problem with the reliability of AI-powered automation. Until OpenAI brings both transparency and real stability, users cannot trust that scheduled tasks will always run. For now, only vigilance and manual checks can guard against silent failures in ChatGPT Work Mode.
