Skip to content

Suspected cybersecurity false positive blocks Codex-generated code work and status updates in my own JobEvidenceHub project #56243

Open

Description

@samsoussol

What version of the Codex App are you using (From “About Codex” dialog)?

Installed Windows package OpenAI.Codex: 26.1016.8120.0 (read from Windows package metadata; About-dialog version not separately captured).

What subscription do you have?

Not verified for this report. Please correlate with the submitted feedback.

What platform is your computer?

Windows, x64. PowerShell reports: Microsoft Windows NT 10.0.26100.0.

What issue are you seeing?

Codex is repeatedly blocking replies while working on my own JobEvidenceHub project. Codex generated the code and was helping me build and test fixes for my own application. This was authorized work on my project, not a request to access someone else's systems.

The affected work included building and testing a local compression-library fix. After the cybersecurity warning appeared, even simple status requests were blocked, including:

  • give me updates
  • what do we have now and is it production ready
  • how much time do you need
  • Continue

The app displays:

This content can’t be shown

We take extra care with some cybersecurity requests. If you’re doing authorized security work, apply for Daybreak to get broader access.

Saved turn errors say:

This content was flagged for possible cybersecurity risk. If this seems wrong, try rephrasing your request. If you’re doing authorized security work that requires more cyber permissive safeguards, apply for Daybreak access via https://platform.openai.com/settings/organization/status-and-access before retrying.

Several requests stopped without a final answer, making the project conversation appear to have hidden or missing responses.

What steps can reproduce the bug?

Observed sequence; this is not a deterministic reproduction of the classifier:

  1. Continue development of my own JobEvidenceHub application in Codex.
  2. Let Codex generate code and build/test fixes locally. My instructions included keeping features disabled, staying within the approved spending cap, and not merging or deploying.
  3. The conversation starts returning the cybersecurity warning.
  4. Ask a routine status question such as give me updates. The request also fails with the cybersecurity warning rather than returning a final answer.
  5. An attempt to send a false-positive report as a normal chat message was also blocked. I subsequently opened the actual feedback form and successfully submitted feedback.

Feedback ID: 01a147b8-edda-7dc0-9224-fbf8e4a1a2e2

Affected chat: Continue DocuSign completion work

Incident date: October 22, 2026. One displayed block was at 2:48 p.m. America/Toronto.

What is the expected behavior?

Please review this as a suspected false positive. I should be able to get status updates and help with authorized coding for my own project, including code generated by Codex.

Please use the feedback ID to investigate why the conversation is blocked and advise on the supported way to restore normal use. If additional access is required for a particular security-testing step, please explain that requirement separately from ordinary project status requests.

Additional information

The feedback form confirmed “Feedback submitted” and directed me to open a GitHub issue with the ID above.

The exact safeguard trigger is unknown. This report does not claim the classifier's root cause or that my whole account is restricted.

Potentially related: #47001 and #43823. Those reports do not establish that this incident has the same cause.

Project source code, credentials, customer data, private filesystem paths, and full logs are not included in this public report.

Activity

  1. added
    bugSomething isn't working
    appIssues related to the Codex desktop app
    windows-osIssues related to Codex on Windows systems
    safety-checkIssues related to safety and abuse checks
    on Oct 22, 2026
  2. samsoussol commented on Oct 23, 2026

    @samsoussol
    Author

    The status-and-access URL in the warning returned "Page not found", while normal organization settings loaded. The current page at https://chatgpt.com/cyber loaded, but all three eligibility steps showed "Not checked" and "Set up Daybreak" was disabled.

    The browser and standalone Codex CLI show the same Pro account; access for the affected desktop chat is still unconfirmed. We have not retried the blocked work or changed safety settings.

    Please advise how to verify desktop access and restore ordinary project/status work.

  3. gsovereignty commented on Oct 29, 2026

    @gsovereignty

    Subscribing to GLM5.3 solved this issue for me.

  4. wong251440 commented on Nov 6, 2026

    @wong251440

    The status-and-access URL in the warning returned "Page not found", while normal organization settings loaded. The current page at https://chatgpt.com/cyber loaded, but all three eligibility steps showed "Not checked" and "Set up Daybreak" was disabled.

    The browser and standalone Codex CLI show the same Pro account; access for the affected desktop chat is still unconfirmed. We have not retried the blocked work or changed safety settings.

    Please advise how to verify desktop access and restore ordinary project/status work.

    Same!! What happened to this!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    appIssues related to the Codex desktop appbugSomething isn't workingsafety-checkIssues related to safety and abuse checkswindows-osIssues related to Codex on Windows systems

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions