AI policy: federal gridlock, state action
White‑house tech advisers pitch a bipartisan path for an AI bill focused on child safety, utility costs and IP, but critics say it leaves vulnerable groups exposed—and states are already moving ahead with their own, fragmented AI rules. The result is a growing patchwork of compliance risks for engineers building AI systems. (siliconvalley.com) (techpolicy.press) (npr.org)
The White House published a four‑page "National Policy Framework for Artificial Intelligence" on March 20, 2026, asking Congress to enact a single federal AI law that would preempt a growing number of state rules. (whitehouse.gov) Senior White House tech advisers, including OSTP director Michael Kratsios and former AI czar David Sacks, have publicly urged Congress to convert that framework into law "in the coming months" to avoid a state patchwork. (npr.org) (knpr.org) The framework calls for child‑safety requirements only on "AI platforms likely to be accessed by minors," proposes a Ratepayer Protection Pledge to shield residential electricity rates from AI data‑center expansion, and pushes for limits on developer liability and federal preemption of stricter state laws. (mayerbrown.com) Policy analysts and civil‑society groups criticized the framework for narrowing protections for marginalized groups and leaving open critical gaps on discrimination, surveillance harms and intellectual‑property enforcement in a March 27 commentary titled "Trump’s AI Policy Framework Leaves Most Vulnerable Exposed." (techpolicy.press) State legislatures have already moved: the National Conference of State Legislatures reported 38 states adopted or enacted roughly 100 AI measures during 2025, while the multistate.ai tracker counted more than 1,500 active AI‑related bills across states as of March 2026. (ncsl.org) NPR documented a White House intervention this year that stalled a Utah transparency bill after a one‑line memo from the administration, a development Rep. Doug Fiefia said kept his transparency measure from receiving a floor vote. (npr.org) (knpr.org) Legal advisories note several state statutes — including measures in California, Texas and Illinois — took effect January 1, 2026, while Colorado's more comprehensive AI Act is scheduled to begin June 30, 2026, creating immediate multi‑jurisdictional compliance dates engineers must track. (bakerbotts.com) The White House has said it wants the framework turned into law "this year," and law firms warn courts will likely be asked to decide how — and whether — a federal preemption push can override dozens of existing state AI statutes. (cnbc.com)
