An AI model must not invent information. All numbers, risks, and problems reported by an AI model must come from real data or clear rules. An AI model must state 'No remaining risk' when no risks exist.
By Pawan Nayar · from the LLOS AI Learning record
An AI model must not invent information.
All numbers, risks, and problems reported by an AI model must come from real data or clear rules. An AI model must state 'No remaining risk' when no risks exist.
An AI model must base every quantity on a measurement or calculation. An AI model must base every concern on a document or a stated rule. An AI model must not create numbers, like 'might use 1-2GB RAM,' if there is no proof. An AI model must not invent risks just to fill a section. An AI model must not blame problems on specific platforms without evidence. If there are no problems, an AI model must clearly state, 'No remaining risk.'
A developer asks an AI model to list potential risks for a new software feature. The AI model lists 'high memory usage on older browsers' without any data. This makes the developer question the AI model's other risk assessments. The developer cannot trust the AI model's reports.
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 must only state facts that it can prove. Do not guess numbers for memory, speed, or other technical details.
An AI model checks user text for certain words. It uses a special list for this check.
Server code contains important safety features. These features protect the system from errors.
When an AI model is unsure about content, it should always score the content as sacred. Sacred content needs special handling.
An AI model must not invent information. All numbers, risks, and problems reported by an AI model must come from real data or clear rules. An AI model must state 'No remaining risk' when no risks exist.