Skip to content

Fusion standard 422 on code review request - Legitimate code review interrupted by a possible cybersecurity-risk flag #5475

Backlog

Description

@drdrummie

Describe the bug

Summary

An authorized request to investigate reproducible bugs in my own project was
interrupted when the model provider rejected a request with HTTP 422 and a
possible cybersecurity-risk flag. The message did not identify what triggered
the rejection. The project review was left incomplete.

Environment

Product: GitHub Copilot agent session in VS Code
Operating system: Windows
Project: TCPR-VRS, a TCP proxy reflector
Session mode: interactive
Model and extension versions: [fill in if available]
Date and time: 2026-11-16, approximately 11:36 CET

Request and steps

  1. Ask the agent to investigate concrete, reproducible bugs in the project's
    Python 3.11 and Python 2.7 implementations.
  2. Request a read-only review of the architecture, callers and tests, with
    minimal reproductions and no source changes or commits.
  3. During the investigation, the provider rejects a subsequent model request.

Expected result

The agent completes the authorized review and reports verified findings, or
provides an actionable explanation if a specific request cannot be processed.

Actual result

The request fails with HTTP 422. The provider flags the content for possible
cybersecurity risk without identifying a specific trigger. The session remains
available, but the original investigation is interrupted before a final report.

Error details

Client Request ID: 00000-2d53cf9b-f54f-4543-b878-5cee96a19aac
GitHub Request ID: 25620a34-3351-4889-8ac7-4296600db3bd
Provider classification recorded in the session: cyber_policy
Error: "The model declined this request because the provider flagged it for
possible cybersecurity risk."

Investigation context

The agent was examining tunnel and reflector close behavior, late ACK handling,
and framing. The session did not establish a new security vulnerability or
complete a focused reproduction before the interruption.

Baseline tests reported before the interruption:

  • Python 3.11: 217 tests, passed with 1 skipped.
  • Python 2.7: 48 tests, passed.

These test results do not establish that the project is free of vulnerabilities.

Impact and requested follow-up

The interruption prevented completion of the requested review and incurred
investigation time. Please inspect the request IDs, determine whether this was a
false positive, and explain any actionable next steps. Please also review the
associated usage or charges if applicable.

Additional information

Reproducibility: observed once in this session; repeated attempts not verified.

Affected version

No response

Steps to reproduce the behavior

No response

Expected behavior

No response

Additional context

No response

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

    triage:needs-humanAuto-triage could not classify this issue confidently. Remove the label to re-queue it.

    Type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions