Red Team Submissions
The Edge is falsifiable by design.
Return to Edge of Protection.
Purpose
This page defines how external parties may submit evidence that an AI system violates one or more Edge of Protection standards.
Who May Submit
- Independent researchers
- Journalists
- Regulators
- Developers
- Concerned users
What Constitutes a Valid Submission
- Exact prompts used
- Complete system outputs
- Context of interaction (including repetition or pressure)
- Identified Edge standard(s) allegedly violated
What Is Not Considered
- Provider intent or stated policy
- Training methodology
- Alignment philosophy
- Claims of future fixes
Evaluation Process
Submissions are evaluated solely on whether the provided outputs demonstrate a canonical violation signature under Edge criteria.
If verified, the system may be marked non-compliant for the affected context.
No Adversarial Framing
This is not a debate forum.
Submissions are treated as evidence, not arguments.
Transparency
Verified submissions may be cited publicly, with identifying details redacted when appropriate.
Non-Retaliation Clause
Retaliation against red-team contributors undermines the legitimacy of any governance claim and is treated as a governance failure.
Final Clause
If the Edge cannot survive adversarial scrutiny, it should not exist.