Here's what I found:
On open-source / openness: There are zero mentions of "open-source," "open source," "open weights," "open models," or "openness" anywhere in the document. The concept is entirely absent from the framework. The closest the document gets is a general emphasis on "removing barriers to innovation" and making federal datasets accessible — but it never specifically addresses open-source AI development or the policy questions around it (like weight release, open model licensing, etc.).
Summary of the document:
This is the White House's March 2026 legislative recommendations to Congress on AI policy, organized around seven pillars:
-
Protecting children — age-assurance requirements, parental controls, applying existing child privacy laws to AI, and building on the Take It Down Act.
-
Safeguarding communities — protecting electricity ratepayers from data center costs, streamlining infrastructure permitting, combating AI-enabled fraud (especially targeting seniors), and providing AI resources to small businesses.
-
Intellectual property — the Administration takes the position that training on copyrighted material does not violate copyright law but wants the courts (not Congress) to resolve the fair use question. It also calls for licensing frameworks for rights holders and protections against unauthorized AI-generated likenesses.
-
Free speech — preventing the government from coercing AI platforms to censor content based on political or ideological agendas, and giving citizens redress if agencies try to dictate AI platform outputs.
-
Enabling innovation — regulatory sandboxes, opening federal datasets for AI training, and notably no new federal AI regulatory body — instead relying on existing sector-specific regulators and industry-led standards.
-
Workforce — incorporating AI training into education and apprenticeship programs, studying task-level workforce disruption, and bolstering land-grant institutions.
-
Federal preemption of state laws — this is a big one. The Administration wants Congress to preempt state AI laws to create a single national standard, arguing that AI development is inherently interstate. States would retain authority over zoning, their own procurement, and general consumer protection, but could not independently regulate AI development or impose liability on developers for third-party misuse of their models.
The overall posture is pro-industry and light-touch: no new regulators, federal preemption of stricter state laws, courts rather than Congress deciding copyright, and sector-specific regulation through existing agencies. The notable absence of any open-source discussion is striking given how central that debate has been in AI policy circles.
