NIST AI governance: standardize now vs keep developing
Risk-based AI governance is ready for immediate standardization.
Security, privacy, transparency, incident response, recovery, training-data practices, terminology, and taxonomy are near-term standards candidates.
TEVV procedures can be standardized now, but TEVV metrics and scientific validity still need development.
Human oversight alone is not enough; NIST points to unresolved measurement and configuration gaps, so standards should also require accountability, documentation, feedback, and independent testing.
Risk and trustworthiness metrics, realistic benchmarks and datasets, interpretability evidence, energy measures, and cross-hardware comparisons still require active development before rigid universal standards.
Create your account to keep this answer and continue from it later.
Let's look at alternatives:
- Modify the query.
- Start a new thread.
- Remove sources (if manually added).