Great American AI Act draft codifies NIST AI center
The sharper move is federal preemption: states would lose frontier-model development authority while keeping the messier post-deployment fights.
TL;DR
FedScoop reports that the bipartisan Great American AI Act discussion draft would codify the National Institute of Standards and Technology’s Center for AI Standards and Innovation and authorize $100 million annually for fiscal 2027 through 2029. It also would preempt state frontier-AI development rules while allowing post-deployment and general-applicability laws. Federal primes, large frontier-model developers and state chief information security officers get a split map: federal standards and incident reporting for frontier development, state oversight after deployment. The missing trigger is practical: which critical safety incidents force reporting to government.
According to FedScoop, the Great American AI Act remains a discussion draft, which matters because it has no formal introduction date and its $100 million annual authorization for fiscal 2027 through 2029 is only the first budget fight. Still, the shape is concrete: codify Commerce’s Center for AI Standards and Innovation, give it AI security standards, evaluation and monitoring work, and turn an executive-branch standards shop into a congressional program.
That pairing is the policy move. Congress would give permanence to an artificial intelligence standards body while also deciding which regulators get the first bite at frontier-model development. The draft would block state laws regulating development of frontier AI models, while preserving state post-deployment regulation and laws of general applicability. Executives get a cleaner development rulebook. State security and procurement officials get the later, harder problem of governing systems after they enter agencies and markets.
For contractors and AI developers, the practical signal is that voluntary standards may become the baseline everyone points to. The draft would require large frontier developers to report critical safety incidents to the government. It also would send the Government Accountability Office into federal AI adoption reviews, establish testbed work involving the Energy Department, the National Institute of Standards and Technology and the National Science Foundation, and authorize Cybersecurity and Infrastructure Security Agency grants for security work by maintainers of widely used open-source software.
The unanswered questions are the ones practitioners need before writing procedures: when the bill will be introduced, whether the $100 million survives markup, and what threshold makes a critical safety incident reportable. Until those definitions land, the Monday work is inventory and dependency mapping. Know where agency AI systems touch frontier-model providers, which state rules would still apply after deployment, and which contracts already reference NIST AI guidance.
Published ·Deep Fathom