Work Mode tools remain unavailable after Sep 14 mitigation: persistent popup and file/media failures

Hello — I am still experiencing a persistent ChatGPT Work Mode tool and workspace failure after the mitigation reported on the OpenAI Status page.

Timeline

  • Work Mode was functioning normally earlier on September 14, 2026.
  • I first noticed the issue at approximately 4:00–6:00 AM PDT (11:00–13:00 UTC).
  • It has continued across multiple Work threads and clients.
  • OpenAI has since posted a related incident stating that some ChatGPT Plus users had limited access to Work Mode workspace tools and files. The status page says a mitigation was applied and the service recovered, with monitoring ongoing:
    Elevated errors affecting Work Mode in ChatGPT - OpenAI Status
  • The issue remains reproducible on my account after that mitigation.

Environment

  • ChatGPT Plus
  • Model: GPT-5.6 Sol, primarily High reasoning and occasionally Extra High
  • ChatGPT web app in Safari on macOS
  • ChatGPT macOS desktop app
  • ChatGPT iOS app used as a comparison
  • Reproduced both inside and outside Projects

Primary symptom: persistent tool warning

Every user turn in a Work thread triggers the following warning:

Some tools are temporarily unavailable

This occurs regardless of whether the input is:

  • Plain text
  • An uploaded image
  • An uploaded video

The warning appears after every turn in both the ChatGPT web app and the macOS desktop app.

The interface does not identify which tools are unavailable, provide any diagnostic details, or offer a way to reload or reinitialize the affected tools.

Scope

I can reproduce the warning in:

  • Existing Work threads
  • Newly created Work threads
  • Work threads inside multiple Projects
  • Work threads outside Projects

Regular Chat threads under the same account do not display the warning.

This makes the issue appear specific to Work Mode rather than to one conversation or Project.

Functional failures discovered during testing

The warning is not merely cosmetic. Several capabilities that previously worked are currently unavailable.

1. Markdown and local file creation fail

I asked the assistant to create and save a minimal Markdown file.

The assistant could still converse and access some connected capabilities, but it did not have a usable local execution/file-writing path from which to create the source .md file. Consequently, no downloadable file was produced.

2. Uploaded PNG files produce a local workspace error

On every PNG uploaded during testing, across at least five consecutive uploads, the message delivered to Codex contained this diagnostic:

Codex could not read the local image at /workspace/scratch/.../upload/<file>.png: No such file or directory (os error 2)

The image nevertheless became visible to the assistant through the separate attachment/vision pathway.

This suggests that the attachment upload completes, but the local workspace file is unavailable when Codex first attempts to read it.

3. Uploaded video files can no longer be analyzed

I uploaded a .MOV file to an affected Work thread. The attachment uploaded successfully, but the assistant had no available capability to open the video, extract frames, play it, or otherwise inspect its contents.

The same account and workflow had previously been able to analyze uploaded videos.

OpenAI’s current documentation states that ChatGPT can accept video-file attachments and may use tools to analyze them, while noting that availability varies and that analysis may be incomplete or inaccurate:

I understand that video analysis is not guaranteed on every platform or account. However, its disappearance coincides with the repeated Work Mode warning, local file-creation failure, and local image-path errors.

iOS comparison

The same Work thread behaves differently in the iOS app:

  • The “Some tools are temporarily unavailable” warning is not displayed.
  • However, the underlying capability failures remain.
  • Markdown creation still fails.
  • Uploaded media still cannot be processed normally.
  • PNG uploads still generate the local workspace No such file or directory diagnostic on the Codex side.

This suggests that the iOS client may not surface the warning even though the underlying Work Mode tool state remains degraded.

Troubleshooting and comparisons performed

I have tested:

  • New and existing Work threads
  • Multiple Projects
  • Work threads outside Projects
  • Web and macOS desktop clients
  • The same thread in the iOS app
  • Plain-text, image, and video inputs
  • Regular Chat threads as a control condition

Changing threads, Projects, clients, and input types does not restore the missing capabilities.

Expected behavior

  • Work Mode tools should initialize consistently.
  • The warning should not appear after every turn.
  • If a tool is unavailable, the interface should identify it and provide a recovery action.
  • A minimal Markdown file should be creatable.
  • Uploaded images should exist at the local path supplied to Codex.
  • Supported video attachments should remain available to the relevant analysis tools.
  • The web, macOS, and iOS clients should accurately reflect the same underlying tool state.

Could this account or its Work sessions still be affected by the September 14 incident despite the reported mitigation?

I can provide affected thread identifiers, exact timestamps, and additional screenshots privately if needed.

Figure 1&2. The Work Mode warning displayed after every user turn, regardless of input type.

Figure 3. Assistant-visible diagnostic from the affected Work thread; the raw ENOENT error occurred on every tested PNG upload.

Same here. While using Work Mode, I received this response:

“This session does not have local file write capability, so it is temporarily unable to generate downloadable files.”

Thanks for confirming - this matches my markdown test too. So the warning clearly isn’t just cosmetic. I’m still seeing it after the mitigation, and the popup is starting to feel like part of the interface… Hope it’ll be fully resolved soon.

Same. Putting in a huge amount of effort into a Project only to have Product Failure is unacceptable for PAID SERVICE! Doesn’t anyone else feel this way? So am I have to spend my $ (ie time is money) fixing their issue?

Same here, exactly. In my case, up until about two hours ago, I was getting the orange error message, but image generation was still working fine.

Then, about an hour ago, GPT-6 started saying “Could not load the image path,” and I could no longer generate images with a reference image. 5.6 had already been left broken for a while, which dragged the image generation quality down too, so lately I’d only been using Work, since that’s where 6 actually works (it eats into my usage quota, but there was no other choice…).

Now even Work has stopped working — I really wish they’d get their act together!

Same here … Was able to work on this just say a couple of hours ago…but now its says excel editing capability was not active

I’ve run into a similar problem too - it’s totally blocking my progress, just like an elephant in the room. No matter what model I use, whenever this issue pops up, it’s always accompanied by the inability to upload PPTX and zip files together :upside_down_face:

Same issues here. Is anyone from OpenAI in on this chat thread?

Hi all,

There seem to be Elevated errors affecting Work Mode in ChatGPT - OpenAI Status. They are right now monitoring the result (according to the status page).

System says both “resolved” and “monitoring”. Isn’t it supposed to be smart? I went through half hour clearing browser cache / moving data to new Work / waiting and it gave me "I can access the required source files, but this Work session currently lacks the local file-processing runtime needed to extract ZIPs, render exact charts, and build/validate new ZIP archives. I won’t fabricate downloads or hashes.

Please retry this request when the file-processing tools are restored; no re-upload should be necessary because the required files are present."

Also, when I asked for compensation for outage, i got “On refunds: I can only process a refund as a subscription refund, and that would cancel Plus—I can’t issue a partial “service quality” credit/refund while keeping the subscription active.” lying and saying “resolved”, no prompt resolution is GARBAGE SERVICE!

I understand your frustation, but I am only another user here sharing the status page.

I know text on a screen can sound worse than it actually is… I just want ChatGPT to hear my (and hopefully others) human voice. Paying for a service should mean the vendor delivers service.

Me sucede lo mismo. Todavía no se solucionó.

This is still an issue as of this message. Tools are temporarily unavailable, and an error along the lines of “Please re-upload the ZIP in a session with workspace access enabled, and I’ll return an updated downloadable solution.”

Same thing happened to me. I thought it was an isolated case.

Yup. This has destroyed Monday for me. Time sensitive project work is now delayed. Please update with plans @OpenAI_Support Thanks.

i received email about case asking for screen recording. of course they’re not paying me to debug for them.

Friends, a small update from my side: the popup has finally stopped making itself part of the interface (・∀・) Image uploads, video processing, and markdown file creation/readback are also working again in my latest tests.

I did encounter one brief local-path error with an image, but the same file worked on retry, so I’m not quite ready to call it fully stable yet. Still, this is the first encouraging sign I’ve seen since the incident. Hope things are improving for everyone here too. Let’s see.

Mine is back to normal too! It didn’t revert at all with version 5.6, so I’m glad it actually went back to how it should be this time.

The images generated right after the fix were of such incredibly high quality that I ended up forgiving GPT lol.

appears to be working. thanks for listening! :slight_smile: