A system message that spells out boundaries and how to refuse in order to reduce harm. It is just one layer in a safety approach rather than a full control on its own.
Read more: Microsoft Learn
In the Ultra Transcenders books
Each book explains Safety system message in context, with comparison tables and the common traps.
Terms in this definition
- Agent instructions
System message of an agent defining its role, objectives, tone and boundaries; it also steers, though not deterministically, when tools get called.
- Full control
Gives every right over protected content, EXTRACT included, plus the ability to alter or strip the encryption. Owners and the Rights Management issuer always hold it.