Desktop ChatGPT Work won't start tasks for one user, even after updating
A user says the desktop ChatGPT Work app throws an error on every new task, even on the latest version. The cause is unknown, and it is a useful reminder to keep a fallback.
The short version
- One user says every new task in desktop ChatGPT Work ends in an error, across all projects.
- The app appears to be on the latest version, so the usual fix of updating does not help.
- OpenAI's status page listed an Oct 7 startup issue, fixed with a patch, and a Nov 11 issue marked resolved.
- The writer reported the bug to OpenAI support and had received no reply at the time of writing.
Why it matters
ChatGPT Work is meant for handing off real jobs like research and documents, so a failure at startup stops the work before it begins. This is one report, not a confirmed widespread outage. Still, it shows why anyone relying on an AI agent needs a fallback route and their files saved outside the app.
A task that never starts
As note reported in a user post, the desktop version of ChatGPT Work has stopped letting its author begin tasks. ChatGPT Work is a feature for handing jobs such as research and document creation to an AI agent, meaning software that carries out a multi-step job rather than just answering a question. According to the writer, an error appears when they try to start a task, and they cannot get any further.
The problem is not tied to a single project. The writer says every project produces the same error, which suggests the trouble is not about the content of any one task.
The writer offers a guess. Judging from the error message, the cloud workspace that ChatGPT Work uses to do its jobs may be failing to get set up. The writer also says they cannot tell whether the fault lies in the app itself or on OpenAI's servers, so this is a theory and not a diagnosis.
What OpenAI's status record shows
The writer checked OpenAI's official status information as of November 15, 2026, and found two entries that looked relevant. The first was an issue on October 7, when new tasks could not be started in the desktop version of ChatGPT Work. OpenAI answered that one with a patched app for Windows, macOS and Linux, and told users to update.
The second was a November 11 issue affecting several features, ChatGPT Work among them. That one is officially marked as resolved.
Here is the snag. The writer's app already appears to be the latest version, so the October advice does not apply. Nothing in the status information describes the error they are seeing now. They therefore cannot say whether it shares a cause with either earlier problem.
The writer sent a bug report to OpenAI support but had heard nothing back when the post was written.
Why a startup failure is its own kind of problem
When an AI agent does a job, several layers have to work in turn. The model makes a plan, tools carry out the steps, a permission check approves actions, and the app ties it all together. This is a general pattern, and nothing in the post says how ChatGPT Work is actually built.
Even so, the pattern helps explain why a startup error stands apart. If nothing begins at all, the trouble is more likely in the app or its setup than in the AI's judgment. The writer's guess about a cloud work environment failing to prepare would fall into that setup stage. That is a reasonable thing to suspect, but it is not confirmed.
The writer's larger point is about trust. ChatGPT Work is meant to get actual work done, not just answer questions, so being unable to start a task is a serious flaw. However much AI features improve, the writer argues, people cannot hand over their work if the tool is not dependable when needed. They want better basic stability and stronger support when something breaks.
If you hit something similar
Our newsletter treated this as a single user's report, and that framing still fits. How many people are affected is not established, and no fix is confirmed. Still, there are sensible steps to take if the same thing happens to you:
- Write down the details. Note the app version and the exact error text. A precise report is much easier for anyone to act on.
- Try the browser version. Check whether the same task runs there while you wait.
- Report it. Use OpenAI's support channels so the problem reaches the team.
- Keep your files elsewhere. Save documents outside the app so a failed install never leaves you without your work.
If you have a deadline, do not let one desktop install be the only route to your work. Keep a second path ready before you need it.
What remains unknown
Several basic questions are still open. Nobody has confirmed whether this is a fault in the app, a problem on OpenAI's servers, or something specific to the writer's setup. It is also unclear whether it connects to the October or November incidents.
The post also shows how hard it can be for users to find answers. The status page did not describe this error, and support had not replied. For people who rely on the tool for deadlines, that gap in information is part of the problem, alongside the error itself.
What to watch
- Whether OpenAI replies to the writer or posts a status entry that matches this startup error.
- Whether other users report the same failure, which would show if this is widespread or one setup.
- Whether a new patched version appears, as it did after the October 7 issue.
Try it
Open your ChatGPT Work setup and note the app version. Confirm that the same kind of task also runs in the browser, and that your key files are saved outside the app.
Sources and links
- Reportingnote: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by note · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
