NIST AI 100-2 Proof
NIST AI 100-2 Proof is Proof of Efficacy for an attack class in the NIST adversarial machine learning taxonomy, produced under PP-SPEC-040 of the Proof Protocol. The taxonomy names attacks and mitigations; Proof Protocol establishes whether the mitigation worked. The name describes a mapping, not an endorsement by NIST.
What it means
NIST AI 100-2 Proof links a framework a reader already trusts to evidence Proof Protocol produces. The framework names what to test. Proof Protocol establishes, through an independent Proof Witness, whether the control worked.
The taxonomy names attacks and mitigations; Proof Protocol establishes whether the mitigation worked.
Where the framework stops short
Frameworks name risks, controls and obligations. Most accept evidence the measured party produced itself: a configuration, a log, a report, an attestation. That is verification, not proof. NIST AI 100-2 Proof replaces it with a measured outcome under adversarial pressure.
What it is not
The name describes a mapping, not an endorsement by NIST. A mapping identifies where a proof may be relevant; it does not establish compliance, certification or endorsement.
Where it is defined
PP-SPEC-040, the NIST AI 100-2 mapping of the Proof Protocol. See also Framework Proof and NIST Proof.