The National Institute of Standards and Technology (NIST) is a U.S. federal agency that has developed a comprehensive program for artificial intelligence safety and trustworthiness. NIST AI encompasses a set of voluntary frameworks, standards, and testing resources designed to help organizations manage risks associated with AI systems. The agency's work in this area gained significant momentum following the 2023 Executive Order on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, which directed NIST to lead federal efforts in AI safety.
NIST's primary contribution is the AI Risk Management Framework (AI RMF), first published in January 2023. This framework provides a structured approach for organizations to identify, assess, and mitigate risks throughout the AI lifecycle. It is organized around four core functions - Govern, Map, Measure, and Manage - and is designed to be voluntary and adaptable across sectors. The framework emphasizes principles such as validity, reliability, safety, security, resilience, accountability, transparency, and fairness.
AI RMF and Related Publications
The AI RMF 1.0 was released on January 26, 2023, following a multi-stakeholder process that included public comments and workshops. NIST subsequently published the AI RMF Generative AI Profile in July 2024, which addresses unique risks posed by Generative AI systems, including Large language models. This profile covers issues such as hallucination, harmful bias, and the potential for misuse. NIST also released companion documents on secure software development for AI and a taxonomy of AI incidents.
In addition to the RMF, NIST has issued several other AI-related publications. The AI RMF Playbook provides practical guidance for implementing the framework. NIST has also developed the Adversarial Machine Learning (AML) taxonomy, published in March 2024, which categorizes attacks on Machine learning systems and provides mitigation strategies. These resources are intended for use by AI developers, deployers, and evaluators across industry, academia, and government.
Testing and Evaluation Programs
NIST operates the AI Test, Evaluation, Validation, and Verification (TEVV) program, which aims to develop metrics and methods for assessing AI system performance and safety. This includes the creation of testbeds and benchmark datasets for evaluating AI capabilities. In 2024, NIST launched the U.S. AI Safety Institute (AISI), established under the Department of Commerce, with NIST providing technical leadership. The AISI conducts research on AI safety, develops testing protocols, and collaborates with international partners.
One notable initiative is the GenAI program, which focuses on evaluating generative AI technologies. This program develops measurement tools for assessing the quality, safety, and trustworthiness of AI-generated content. NIST also participates in the International AI Safety Report, a global effort to synthesize scientific knowledge on AI risks.
Standards and International Collaboration
NIST plays a key role in developing international AI standards through its engagement with organizations such as the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC). NIST contributes to the ISO/IEC 42001 standard for AI management systems and the ISO/IEC 23894 standard on AI risk management. The agency also works with the National Cybersecurity Center of Excellence (NCOE) to integrate AI safety into broader cybersecurity practices.
NIST's AI work is informed by its broader research in Artificial intelligence and Deep learning. The agency's Information Technology Laboratory conducts foundational research in areas such as computer vision, natural language processing, and Neural network robustness. This research feeds into the development of practical guidelines and test methods.
Challenges and Future Directions
NIST AI faces ongoing challenges in keeping pace with rapid technological advances. The voluntary nature of the AI RMF means adoption is not mandatory, though federal agencies are increasingly required to align with it. NIST continues to update its guidance to address emerging issues such as AI-enabled cyberattacks, privacy risks, and the environmental impact of AI training. As of 2025, NIST is developing additional profiles for specific sectors, including healthcare and finance, and expanding its testing capabilities for frontier AI models.
The agency's work is complemented by collaborations with academic institutions like MIT CSAIL and Stanford AI Lab, as well as industry partners. NIST also engages with international bodies to harmonize AI safety standards globally, recognizing that AI systems often operate across borders.
Impact on AI Ecosystem
NIST AI frameworks have influenced how organizations approach AI governance. Major technology companies, including OpenAI, Anthropic, and Google DeepMind, have referenced the AI RMF in their safety practices. The framework has also been adopted by government agencies and international partners, such as the European Union's AI Act, which draws on similar risk-based principles. NIST's emphasis on measurable, testable criteria has helped shift the AI industry toward more rigorous evaluation practices.
Despite its voluntary nature, NIST AI has become a de facto reference point for AI safety discussions in the United States and beyond. Its publications are widely cited in academic literature and industry reports, and its testing programs provide a neutral ground for evaluating AI systems. As AI capabilities continue to evolve, NIST's role in defining safety standards is likely to expand, particularly in areas like autonomous systems and AI-driven decision-making.