The US government released its national AI plan on March 20. And regardless of where you stand politically — the framework is remarkable. Not necessarily for what’s in it. But for what’s missing.
Seven Pillars
The plan rests on seven pillars: protecting children, safeguarding communities, intellectual property rights, preventing censorship, innovation and AI dominance, workforce education, and — the most interesting one — a unified federal framework that would preempt state laws.
That last point is the real story here. The message to Congress is clear: we want a single, innovation-friendly rulebook. Not 50 different laws in 50 states.
Light-Touch Regulation
The approach is deliberately lean. The White House wants “light-touch regulation” — minimal bureaucracy, maximum room for companies. Developer liability would be limited. Regulatory sandboxes would let companies test AI applications in controlled environments.
Sounds great for the industry. Less regulation means faster innovation. But there’s pushback too.
What’s Missing
Here’s what stands out: the plan says very little about how AI systems should actually be governed in practice. Bias audits get a mention, but the details are vague. Questions about deepfakes, disinformation, and autonomous systems are barely touched.
And then there’s the irony: the same administration that’s suing Anthropic over safety concerns just released an AI plan that explicitly calls for less regulation.
What It Means
For the AI industry, this is a positive signal. A unified federal framework would replace the patchwork of state laws — something many companies have been asking for. For European observers, it’s an interesting contrast to the EU AI Act, which puts much more emphasis on control and compliance.
The real question: will Congress actually deliver? Michael Kratsios and David Sacks are pushing for legislation this year. History tells us that’s ambitious.
Sources: