Overview
Severity: MEDIUM | Affected: NIST | Category: policy
The U.S. National Institute of Standards and Technology (NIST) has officially released the AI Red-Teaming Interoperability Framework (ARTIF) 1.0, establishing a standardized methodology for assessing the security and safety of AI models. This new framework, developed in collaboration with industry partners like Microsoft, Anthropic, and the AI Safety Institute, provides a common language and set of procedures for conducting and reporting on AI red-teaming engagements. ARTIF 1.0 introduces standardized threat taxonomies, testing protocols for specific attack vectors like prompt injection and data poisoning, and a formal scoring system for model resilience. The goal is to enable organizations to compare red-teaming results across different models and third-party assessors consistently. Adoption of the framework is currently voluntary, but it is expected to become a de facto requirement for U.S. government contractors deploying high-risk AI systems.