Region — US profile
Model risk guidance, read against models that act.
No federal AI statute, and no shortage of obligation: supervisory guidance written for models that produce numbers, now applied to systems that take actions.
What applies together
- SR 11-7
- Model inventory, provenance, ongoing monitoring and effective challenge. The column that earns it is whether an inventory entry was observed from a runtime or typed into a form.
- NIST AI RMF
- Voluntary, which is exactly why producing the evidence is a signal: nobody is compelled to.
- State law, and it is moving
- Not mapped. Colorado, California and others are legislating on automated decisions at different speeds and in different shapes. A clause map is data, so one can be authored without waiting for us.
Validation is not ours. SR 11-7's core demand is independent validation by people with standing to challenge. The platform records a model's validation status and who set it; it does not validate a model, and a vendor claiming to satisfy SR 11-7 by shipping software has misread the guidance.
What resolving a profile actually changes
A market profile is not a locale setting. Changing it changes what the platform is permitted to do, what it looks for, what it scores you against and how long it keeps a record — and moving a workspace between profiles is itself a governed act with a name against it.
And a workload can be under more than one at once. The profile resolves per decision — narrowest wins — and is sealed into the record, so a pack states which laws this deployment believed applied at the moment the decision was made. How that works →