Key Takeaways
- NIST has renamed and broadened the AI Safety Institute Consortium into the NIST Artificial Intelligence Consortium, shifting focus toward AI measurement, innovation, and adoption.
- The consortium will operate through six specialized task groups covering testing, evaluation, risk annotation, bias mitigation, documentation, and chemical‑biological security.
- New members are invited via a first‑come, first‑served letter‑of‑interest process; existing members must sign an amendment to continue participation.
- Participation entails entering a Cooperative Research and Development Agreement (CRADA) with NIST to collaborate on standards, metrics, and best‑practice tools for trustworthy AI.
- The initiative aligns with NIST’s Strategy for American Technology Leadership, the National Artificial Intelligence Initiative Act of 2020, Executive Order 14179 (2025), and the America’s AI Action Plan.
Background and Original Formation
In 2023, the National Institute of Standards and Technology launched the AI Safety Institute Consortium (AISIC) to unite academia, industry, and government around science‑based guidelines for AI measurement. More than 280 organizations joined the effort, laying groundwork for global AI metrology by developing empirical standards and best practices. The consortium’s early work centered on safety‑related measurement techniques, aiming to create a reliable foundation for evaluating AI systems across sectors. This initial phase demonstrated the value of a collaborative, stakeholder‑driven approach to addressing nascent AI risks while fostering innovation.
Renaming and Expanded Mission
Recognizing the evolving landscape of AI technologies, NIST has rebranded AISIC as the NIST Artificial Intelligence Consortium. The new name reflects an expanded mandate that goes beyond safety to encompass measurement, innovation, and widespread adoption of AI. While the consortium will retain some of its original safety‑focused activities, its primary thrust now includes building an AI evaluation ecosystem, investing in AI‑enabled scientific discovery, and promoting the use of domestically developed AI systems. This shift aims to create a more holistic framework that supports both the trustworthy deployment and the competitive advancement of AI in the United States.
Alignment with National Strategies
The consortium’s reorientation directly supports several federal directives. It implements NIST’s Strategy for American Technology Leadership in the 21st Century, which seeks to accelerate critical and emerging technologies from lab to market through close industry partnership. Additionally, the effort fulfills requirements of the National Artificial Intelligence Initiative Act of 2020, complies with Executive Order 14179 issued in 2025, and advances the goals outlined in the America’s AI Action Plan. By aligning with these policies, the consortium helps ensure that AI development proceeds in a manner that is scientifically rigorous, economically beneficial, and nationally secure.
Task Group Structure – Part I
To execute its broadened mission, the consortium has organized six task groups, each targeting a distinct aspect of AI measurement and application. The AI Testing, Evaluation, Verification and Validation (AI TEVV) Zero Draft Task Group will produce preliminary, stakeholder‑driven drafts of standards that assess whether AI systems meet design requirements and are fit for intended use. These “zero drafts” serve as comprehensive foundations for subsequent private‑sector standardization processes. Simultaneously, the Annotation for AI Risks & Validity Task Group is developing a science‑based toolkit for evaluating AI risks and impacts, tailored for use with NIST’s ARIA (AI Risk Impact Assessment) program.
Task Group Structure – Part II
The AI Evaluation and Measurement Methods Task Group will survey the field to identify gaps, barriers, and open questions in AI evaluation science, incorporating input from diverse organizations, sectors, and stakeholder roles. Addressing trustworthiness concerns, the Bias Effects and Notable Generative AI Limitations (BENGAL) Group partners with IARPA to devise scalable solutions for misinformation, sensitive‑information leakage, flawed reasoning, and susceptibility to attack in large language models, aiming to enable confident use of these models for intelligence analysis. Lastly, the AI Documentation Cards Task Group will create standardized, practical templates for documenting datasets, models, systems, and TEVV processes, drawing on existing best practices and community feedback to improve transparency and reproducibility.
Chemical and Biological Security Task Group
In addition to the six primary groups, NIST is reviving the Chemical and Biological Security Task Group from the consortium’s earlier incarnation. This group will focus on sharing insights about emerging AI measurement and evaluation techniques that are relevant to chemical and biological threats. By examining how AI can be leveraged—or potentially misused—in these security domains, the task group aims to develop metrics and guidelines that help safeguard critical infrastructure while fostering responsible AI innovation. Its revival underscores the consortium’s commitment to addressing cross‑disciplinary risks that intersect with national security priorities.
Membership Invitation and Process
NIST is openly soliciting letters of interest from any organization possessing the technical expertise needed to contribute to the consortium’s objectives. Submissions will be reviewed on a first‑come, first‑served basis, with selected participants invited to join the consortium. Existing members from the original AISIC are not required to reapply; however, they must sign an amendment acknowledging the updated scope and governance changes. This approach balances inclusivity with continuity, allowing new voices to enter while preserving the foundational knowledge of current partners. Detailed instructions for submitting a letter of interest are available on the consortium’s project website, and further guidance is provided in the associated Federal Register notice.
Cooperative Research and Development Agreement (CRADA) and Implications
Upon acceptance, member organizations will enter into a Cooperative Research and Development Agreement (CRADA) with NIST. The CRADA framework facilitates joint research, sharing of proprietary data (under protected conditions), and collaborative development of standards, tools, and best practices. Through these agreements, the consortium aims to accelerate the creation of reliable, scalable, and interoperable AI metrics that can be adopted across industry, academia, and government. Ultimately, the expanded NIST Artificial Intelligence Consortium seeks to strengthen U.S. leadership in AI by grounding innovation in rigorous measurement, fostering trustworthy deployment, and ensuring that advances benefit both economic competitiveness and national security.