Model version
A new model can behave differently on the same inputs.
Models are upgraded, prompts are edited, tools are added. NVX is designed to show which approvals and evidence are affected by a change, so reviews happen when they are needed rather than once a year.
What goes stale
A new model can behave differently on the same inputs.
Small wording changes can alter outputs and safety behaviour.
New tools extend what an agent can do.
New or changed data affects accuracy, privacy and fairness.
How NVX approaches it
Versioned assurance records are planned for the private MVP. The goal is that each approval references the exact configuration it covered, and a later change marks the affected evidence for review.
01
Capture model, prompt, tool and data configuration at approval.
02
Register a new version when any of those change.
03
Show which controls and evidence relied on the previous version.
04
Route the system back to its owner for a recorded decision.
Questions
NVX is accepting selected enterprise teams into a private preview of its AI governance and assurance workflow.