HOW CONGRESS CAN BUILD OFF TRUMP’S AI FRAMEWORK THIS TAX SEASON

That is why the White House’s national AI legislative framework arrives at such a consequential moment. Its release does not replace the need for legislative action or the debate required to align on a durable federal framework. But it does reflect a growing consensus in Washington that AI is too economically and strategically significant to leave unaddressed. The central question now is not whether rules will emerge, but who will write them and on what terms. If Congress does not act, states, federal agencies, and foreign governments will continue to move ahead on separate tracks, increasing the likelihood of a fragmented and unwieldy regulatory environment.

As Congress considers how to respond to AI, the question is not simply whether to act, but how to act wisely. The goal should be to address real risks without constraining the many AI applications already delivering clear economic and societal benefits. Congress should resist broad, fear-driven regulation that treats every model or use case as equally risky. A better approach is a targeted federal framework, one that focuses on the areas where harm is most likely and guardrails are truly needed while preserving space for beneficial and low-risk innovation.

That framework should begin with a simple principle: Safeguards should scale with risk. Not every AI model deserves the same level of scrutiny. Congress needs to distinguish between systems that draft emails or optimize supply chains and systems that influence critical decisions in health care, infrastructure, defense, or national security.

The most advanced models, particularly those with dual-use national security implications, should face stronger expectations around evaluation, testing, and security planning. Developers of those systems should maintain clear security frameworks that explain how capabilities are assessed, how risks are monitored, and how safeguards evolve as systems become more capable.

Congress already has the right institution to anchor that work: the National Institute of Standards and Technology.

NIST’s AI Risk Management Framework is the most credible technical foundation for trustworthy AI governance because it reflects science, measurement, and operational practice rather than political theory. Congress will not write technical requirements fast enough to keep pace with frontier systems, but it can establish durable statutory principles while directing NIST to develop the testing methods, measurement tools, and trust standards that agencies and industry can apply consistently.

A federal framework should focus on uses rather than the underlying technology itself. AI will affect healthcare differently than transportation, education, finance, or manufacturing. National legislation should establish baseline expectations while allowing sector-specific agencies to tailor them where appropriate. That balance matters. It protects the public while limiting regulatory overreach and preserving the flexibility needed for sector-specific expertise.

Finally, any federal framework should include baseline standards for model transparency. Clear expectations can build trust and help identify surface risks earlier. Companies should document how advanced systems are trained, evaluated, and monitored, particularly agentic systems that can access external data or take actions with limited human intervention. Practical tools such as model cards can help communicate system capabilities, limitations, and intended uses without forcing disclosure of proprietary model weights or trade secrets.

That kind of clarity matters because public confidence in AI will not come from slogans about innovation. It will come from demonstrating that systems are tested, risks are understood, and safeguards are real. Americans want AI to improve their lives, but they also want assurance that automated systems do not amplify fraud, and advanced models do not create new vulnerabilities in critical systems.

The same clarity matters for economic competitiveness. Companies cannot build nationally deployed AI systems under constantly diverging expectations and still move at the speed global competition now demands. A federal framework should provide standards that give innovators confidence to invest, consumers’ confidence to adopt, and policymakers confidence that essential protections are being applied coherently.

There is also a rare political opening here. AI remains one of the few policy areas where both parties agree federal action is necessary, even if they emphasize different concerns. 

TRUMP’S AI FRAMEWORK IS A GOOD START. HERE’S WHAT’S MISSING TO PROTECT CHILDREN

That opening may not last. Every month Congress delays, more states legislate independently, more businesses adapt to conflicting obligations, and more global competitors shape standards the United States may later be forced to accept rather than write.

The White House has now placed federal AI legislation squarely on the congressional agenda. The real question is whether lawmakers will use this moment to build a durable national standard or risk ceding leadership in the most consequential general-purpose technology in decades to foreign competitors. AI is moving too quickly, and the stakes are now too high, for Congress to keep watching from the sidelines.

Liz O’Bagy is Director of Federal Policy, AI Policy Lead for TechNet, where she helps drive TechNet’s federal policy advocacy on key priorities including artificial intelligence, trade, modernizing government technology, the future of work, and more.