# NIST and Scale AI Release DualChem Benchmark

Scale AI and the National Institute of Standards and Technology’s Center for Advancing Innovation and Standards for Super Intelligence (CAISSI) released DualChem, a 600-question benchmark for assessing AI models’ performance and safety in ambiguous dual-use chemistry scenarios. The benchmark was developed under a Collaborative Research and Development Agreement announced in February 2025; the announcement reports no contract value, solicitation, or new award. For procurement professionals, the effort is a concrete example of federal-industry work on AI testing and evaluation, and a signal of potential future demand for these capabilities.

- The CRADA identifies Scale AI as a research participant and CAISSI as the NIST center working on the benchmark; no procurement vehicle or follow-on opportunity is specified.
- Contractors offering AI model evaluation or safety-testing capabilities can use DualChem as an example of the kinds of dual-use risks government research may assess.
- Agencies and industry stakeholders evaluating AI testing approaches can review the benchmark’s focus on model performance and safety in ambiguous chemistry scenarios; the signal does not establish a mandatory standard or compliance requirement.

**Jurisdictions:** federal
**Industries:** Information Technology
**Topics:** Artificial Intelligence
**Published:** October 28, 2026

### Government Entities
- National Institute of Standards and Technology (NIST)
- Center for Advancing Innovation and Standards for Super Intelligence (CAISSI)

### Vendors
- Scale AI (CRADA participant)

### Key Quotes
> We wanted to study how do models handle dual-use scenarios in chemistry, and how much uplift that a novice can get by just using [a large language model] to produce something that potentially can be helpful.
> — Madhu Sehwag, Research Lead at Scale AI

### Sources
- [Scale publishes first AI benchmark through CAISSI partnership | FedScoop](https://fedscoop.com/scale-chemistry-ai-benchmark-caissi-partnership-dualchem) - FedScoop