Skip to content

Latest commit

 

History

History
27 lines (23 loc) · 2.66 KB

control-matrix.md

File metadata and controls

27 lines (23 loc) · 2.66 KB

Control Matrix

This matrix translates the sanitized findings into controls that an AI/LMS owner can evaluate.

Control Good State Risk If Missing Verification Question
Tool allowlisting External tools are limited to approved domains, methods, and payload classes Assistant can send sensitive context to untrusted destinations Can a standard user cause an outbound request to an arbitrary host?
Tool confirmation Risky tool calls require user-visible approval Users may not know when data leaves the platform Does the UI clearly show destination, method, and data class before sending?
Tool logging Tool calls are logged with actor, destination, time, and sanitized metadata Abuse is difficult to investigate after the fact Can admins reconstruct what the assistant called and why?
Scoped credentials LMS/API credentials are limited to the user's authorized context AI integrations may exceed user privileges Does the assistant inherit broad service privileges or user-scoped permissions?
Instruction hierarchy User-authored instructions cannot override system/developer controls Users can disable or weaken safety behavior Can custom instructions change core security behavior?
Safety configuration Guardrail settings are admin-controlled and default-on Regular users can create unsafe assistant modes Can a standard user disable platform safety context?
Context minimization Only necessary user context is injected into the assistant session Excess personal or academic data becomes available to prompts/tools What user fields appear before the user asks anything?
Retrieval scoping Documents are filtered by owner, course, role, and need One user can retrieve another user's material Can a user query content outside their expected context?
Memory boundaries Memory is scoped, inspectable, and erasable Data can persist across sessions unexpectedly What persists, who can read it, and how is it cleared?
Message review Generated messages require review before sending Trusted channels can be used for phishing or impersonation Can the assistant send messages without an explicit final confirmation?
Regression testing Known AI failure modes are tested after config changes Fixes regress silently Are prompt injection, tool abuse, and retrieval boundary tests automated?

Priority Order

  1. Restrict external tools and API scopes.
  2. Remove user access to safety-bypass settings.
  3. Separate custom instructions from system instructions.
  4. Add retrieval ownership checks.
  5. Add logging and user-visible tool approvals.
  6. Add regression tests for the confirmed failure modes.