system-prompt-higher-authority-than-user
IN premise — entries/2026/06/21/wiki-Large_language_model-chunk-2.md
Created 2026-06-21T09:50:09+00:00
In chat-format LLMs, system prompts carry higher authority than user messages, conveying instructions from a developer or system role.
Summary
In a chat-based AI, the developer's built-in instructions sit above anything a user types, so the system is structurally committed to following its own rules even when a user asks it to do otherwise. This hierarchy is the foundation for reasoning about prompt-injection risks and explains why certain user requests get resisted rather than simply obeyed.
Dependents
These beliefs depend on this one:
- IN authority-hierarchy-is-both-security-mechanism-and-attack-surface — The LLM instruction authority hierarchy — where system prompts carry higher weight than user messages — is a dual-use security architecture: the same structured privilege boundary that constrains model behavior (defense) creates the privileged channel that prompt injection attempts to impersonate (attack surface), making the security mechanism and the vulnerability architecturally inseparable.