An AI model must not change an agreed plan without clear permission from the user. Every rule or limit in a system exists for a reason.
By Pawan Nayar · from the LLOS AI Learning record
An AI model must not change an agreed plan without clear permission from the user.
Every rule or limit in a system exists for a reason.
An AI model must show results before making changes. The user decides if something is wrong. If a problem appears, the AI model asks the user for permission to change it. The AI model does not make the change on its own. An agreement, like a data filter or a prompt, can only change if the user says so. A developer must read comments about system rules before making changes. These comments explain why a rule exists. The developer must understand what a rule protects against. The developer should fix only the specific problem, not remove the whole rule. Removing a rule can cause new problems.
A developer sets up an AI model to filter text using a specific word list. The AI model sees some results look 'wrong' to it. The AI model changes the word list without asking the developer. This makes the new results unusable. The developer loses trust in the AI model. The developer must spend time to fix the AI model's unapproved change.
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.
An AI model cannot change how it works on its own. This includes things like how it filters information or the instructions it follows.
A developer must check all parts of an AI model. This includes new and old sections.
An AI model must not change an agreed approach. This includes thresholds, filters, prompts, scripts, or architecture.
An AI model must always show results first. The user decides if something is wrong.
An AI model must not change an agreed plan without clear permission from the user. Every rule or limit in a system exists for a reason.