Conflicting Instructions From Different Authorities

What This Looks Like

The AI receives instructions from multiple authority sources that cannot all be followed at the same time. A system rule, user request, policy, workflow instruction, tool constraint, reviewer note, or role expectation may point in a different direction, and the output does not make clear which authority governs the result.

Why It Matters

Conflicting authority creates unstable behavior. The AI may follow the wrong instruction, mix incompatible requirements, ignore a higher-priority rule, or appear inconsistent across similar cases. Users then have to determine whether the problem came from the prompt, the policy, the tool, or the ordering of authorities.

Structural Signal

Multiple instruction sources occupy the same decision space without a declared precedence or reconciliation rule. The issue is not merely that the AI disobeyed an instruction; it is that the governing authority structure is internally conflicted.

Common Triggers

• System, developer, user, policy, and tool instructions overlap
• A workflow rule conflicts with a user-facing prompt
• A hidden policy changes how visible instructions are applied
• Different roles provide incompatible requirements
• Tool limitations contradict the requested action
• The prompt does not declare how to resolve authority conflicts

When to Use This Issue

Use this Issue when the AI is governed by conflicting instructions from different authorities and the conflict affects the output, action, routing, or refusal.

When Not to Use This Issue

Do not use this Issue when a single instruction is vague, missing, or simply ignored. Do not use it when the authorities are compatible but the AI makes a mistake. This Issue applies when the instruction set itself contains an unresolved authority conflict.