Codex desktop: reviewer safety block without visible findings or recovery controls

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

26.903.8094.0 (installed Windows package version)

What subscription do you have?

Chat GPT PRO

What platform is your computer?

Microsoft Windows NT 10.0.26200.0 x64

What issue are you seeing?

During a GPT-6 Astra-led software implementation on Windows, reviewer verification_boundary_final_review reported: "This request was blocked by our safety systems. Reason: Potentially unintended activity."

I cannot see the associated safety notice, findings, or recovery controls. The parent task subsequently completed normally. Unrelated isolated fixes were later completed, so I am not claiming that all coding is blocked.

Please investigate the missing owner-visible notice and explain the supported way to review the findings and safely continue the affected work. I am not requesting a safeguard bypass.

What steps can reproduce the bug?

Observed sequence on 2026-09-10, not a deterministic minimal reproduction:

1. An existing implementation task reached an independent review.
2. The reviewer reported the safety block quoted above.
3. The parent checkpointed and ended; no findings/resume control was visible to me.
4. I submitted in-app feedback and received the session ID below.

No blocked operation was retried to prepare this report.

What is the expected behavior?

Show the affected task's available safety findings and a supported recovery or escalation path, while preserving safeguards and completed work.

Additional information

In-app feedback/session ID: 019fed18-37b3-7942-9ee5-a604d914c57f.
Model: GPT-6 Astra.
The app version was read from the installed Windows package, not the About dialog. No logs, screenshots, credentials, or private project files are attached. Please offer a secure channel if additional diagnostics are needed.

Start with the reported Codex desktop behavior on Windows and the supplied session ID, 019fed18-37b3-7942-9ee5-a604d914c57f; the report provides no source file, test, logs, or deterministic reproduction. Done means identifying why the owner-visible safety notice, findings, and supported recovery or escalation path were not shown while preserving safeguards.