An AI model's internal rules must be kept separate from its output. The AI model must never show or talk about its own rules. The AI model's output should only be the requested information.
By Pawan Nayar · from the LLOS AI Learning record
An AI model's internal rules must be kept separate from its output.
The AI model must never show or talk about its own rules. The AI model's output should only be the requested information.
A developer must separate the AI model's core rules from its output template. The developer should place all strict rules at the end of the template. These rules need a clear label that tells the AI model not to output them. The AI model's response should only contain the requested content.
A developer creates an AI model to write short sayings. The developer puts instructions like 'do not use long words' inside the main text area. The AI model then includes 'do not use long words' in its output. The developer moves this instruction to a separate, hidden section. The AI model now only outputs the short saying.
Resolved from this lesson's own record — real definitions, not invented for this page.
More from the same record, pulled by shared topic — not written new for this page.
A developer should put absolute constraints at the very end of a prompt. These constraints are golden rules.
A single, very long prompt tries to cover too many situations. This makes the prompt complicated and hard to manage.
Proprietary information includes unique rules, prompt templates, and internal voice descriptions. This information can leak if an AI model outputs its own instructions.
An AI model often has strong ideas about famous authors. These ideas can override general style instructions.
An AI model's internal rules must be kept separate from its output. The AI model must never show or talk about its own rules. The AI model's output should only be the requested information.