AI Threat Administration Framework. Launched on Jan. 26, 2023, NIST’s AI RMF was developed to higher handle dangers to people, organizations, and society related to AI. “What we’re making an attempt to do with the AI Threat Administration Framework is perceive how we belief AI, which operates in some ways otherwise in a few of these duties that we all know very effectively,” significantly concerning how high-impact functions have an effect on cybersecurity, Martin Stanley, principal researcher for AI and cybersecurity at NIST, mentioned on the workshop.
Middle for AI Requirements and Innovation (CAISI). NIST’s CAISI serves because the “business’s major level of contact inside the US authorities to facilitate testing and collaborative analysis associated to harnessing and securing the potential of economic AI techniques,” mentioned Maia Hamin, a technical workers member of CAISI, the middle that develops greatest practices and requirements for enhancing AI security and collaboration. It additionally “leads evaluations and assessments of US and adversary AI techniques, together with adoption of international fashions, potential security vulnerabilities, or potential for international affect,” she informed workshop attendees.
NIST AI 100-2 E2025, Adversarial Machine Studying: A Taxonomy and Terminology of Attacks and Mitigations. This NIST report, printed in March 2025,gives a taxonomy of ideas and defines terminology within the area of adversarial machine studying (AML). “Adversarial machine studying or adversarial AI is the sphere that research assaults on AI techniques that exploit the statistical and data-driven nature of this know-how,” NIST analysis crew supervisor Apostol Vassilev mentioned on the workshop. “Hijacking, immediate injection, oblique immediate injection, information poisoning, all this stuff are a part of the sphere of examine of adversarial AI,” he clarified.
