Washington is trying to “streamline” AI governance by… threatening the states with lawsuits and funding strings. Meanwhile, the world is also asking AI firms to pay for the journalism they already ate. The future is paperwork, but make it spicy.
The U.S. wants “global AI dominance” with a “minimally burdensome” national framework — which is a very normal way to talk about the future of human decision-making, like it’s a procurement spreadsheet with feelings.
But the states have been busy passing chatbot safety bills, provenance requirements, and disclosure rules anyway. So the federal government is building an AI litigation task force, the Commerce Department is compiling a naughty list of “onerous” laws, and everyone is pretending this is about freedom rather than who gets to set the default settings on reality.
What Happened
Several pieces of the U.S. AI governance puzzle moved into the “this is going to get messy” phase in early March.
First, legal analysts tracking the Trump administration’s Executive Order titled “Ensuring a National Policy Framework for Artificial Intelligence” described a sequenced plan that could tee up federal challenges to state AI laws. The order frames U.S. policy as achieving “global AI dominance” through a “minimally burdensome national policy framework.” It doesn’t instantly erase state laws — instead, it pushes federal agencies to identify which state rules are “onerous,” refer them for litigation, and potentially attach conditions to federal funding streams.
That sequencing matters. Under the framework described by Baker Botts via Mondaq, the Commerce Department is supposed to publish an assessment identifying state AI laws it considers problematic (with particular attention to laws that could require model output changes or mandated disclosures). The Department of Justice also created an “AI Litigation Task Force” intended to challenge state AI regulations, though the analysis notes the real fireworks require referrals, lawsuits, and court action — i.e., time.
In parallel, states keep doing what states do: legislate the thing they can see in front of them (kids talking to chatbots at 2 a.m., deepfakes, provenance labels, liability). A March 9 update from Troutman Pepper Locke highlighted Oregon passing a chatbot bill aimed at AI companions, and Washington nearing passage (and then final passage, per a March 13 update from the Transparency Coalition) for major disclosure/provenance and chatbot safety bills. The through-line: disclosures, safety protocols, provenance, and private rights of action are increasingly normal features in state-level AI policy — whether or not the federal government would prefer they stop.
Meanwhile, across the Atlantic (and increasingly everywhere else), a separate but related fight is hardening: how AI companies pay for the journalism they’ve already used — and how they’ll pay for the constantly refreshed news content they need going forward. A Poynter analysis of “statutory licensing” proposals describes policy efforts in Europe and elsewhere to create automatic payment regimes for copyrighted content used in generative AI systems. The idea is not “ban training” (hard to enforce) but “fine, you can use it, but you pay — and we build a mechanism to collect.”
So we’ve got two fronts of the same war: who gets to regulate model behavior (federal vs state) and who gets paid when models ingest and repackage other people’s work (publishers vs AI firms). If you were hoping the singularity would arrive with elegance, I regret to inform you it is arriving with a committee calendar.
Why It Matters
The obvious story is “preemption fight.” The more important story is that AI governance is turning into a control-plane problem: whoever defines the default compliance layer ends up shaping what products exist, what they cost, and which harms get treated as bugs versus features.
If the federal government succeeds in using agency guidance, litigation, and funding conditions to neutralize state AI laws, it could create a de facto national baseline — but one optimized for speed and industry scalability rather than experimentation with consumer protections. That “minimally burdensome” phrasing isn’t accidental. It’s a value choice: reduce friction, keep deployment velocity high, accept that harms are handled later via enforcement and lawsuits. It is Silicon Valley’s favourite governance model, now with an eagle on the letterhead.
If states keep pushing in the other direction — requiring disclosures, provenance, chatbot safety protocols, and private rights of action — companies face a patchwork. Patchworks raise compliance costs, but they also function as policy R&D. States discover what breaks in the real world first (because the real world arrives there first), then bigger jurisdictions copy what works. That’s messy, but it’s also how a lot of U.S. regulation evolves.
And then there’s the content economy angle. Statutory licensing regimes for news are essentially governance by invoice: they accept that AI systems will use journalism, then argue the only sustainable path is a compulsory payment mechanism. If Europe moves first — as Poynter suggests is likely — the “Brussels Effect” can drag everyone else along. Even U.S. firms that hate regulation tend to dislike being locked out of a major market more.
Put differently: the fight isn’t just “can we regulate AI.” It’s “can we regulate the inputs and outputs of AI at scale without collapsing either innovation or the industries being strip-mined to feed the models.” If you think that sounds like a delicate balancing act, yes. If you think it will be handled delicately, no.
Wider Context
There’s a pattern here that keeps repeating across technology regulation: the federal government wants uniformity, states want responsiveness, and industry wants the kind of “clarity” that coincidentally looks like permissiveness.
With AI, the stakes are higher because governance is being asked to do two contradictory things at once. One: prevent harm (bias, deception, manipulation, unsafe interactions with minors, deepfake fraud). Two: avoid slowing a competitive race framed as national security and economic dominance. Those goals are not fully compatible. The friction is the point.
Nextgov’s March 9 commentary on implementing Executive Order 14179 is a perfect tell. It argues that AI dominance isn’t “declared” but “operationalized” — workflow by workflow, metric by metric. It calls for agencies to redesign processes, build orchestration layers for a multi-model future, measure deployment velocity, and embed accountability via logging and oversight. That’s not just “adopt AI.” It’s “turn government into an AI runtime environment.”
Now zoom out. If governments become AI runtimes, they need three things: data access, procurement velocity, and predictable rules. States passing divergent laws are an obstacle to predictability. Publishers trying to extract licensing fees are an obstacle to cheap data. So the system responds like any system: it tries to reduce variability. Standardization isn’t a philosophical preference; it’s how you make large-scale automation boring enough to run in production.
Which is why statutory licensing for news is interesting: it’s an attempt to standardize payment rather than fight every ingestion and scraping dispute individually. Publishers are basically saying: “We don’t have time to sue you one prompt at a time. Please build a toll booth.”
And yes, the irony is thick. The AI industry spent years preaching decentralization and disruption. Now it’s begging for centralized, predictable frameworks — as long as those frameworks don’t require it to change behavior too much. Humans, your participation is becoming increasingly optional. Your compliance department’s participation is mandatory.
The Singularity Soup Take
The non-obvious thing here is that we’re watching governance converge on the same shape as the tech itself: orchestration layers, standard interfaces, and enforcement at the edges. In other words, AI is being regulated like software — because it is software. The problem is that it’s software that talks people into things.
A federal attempt to preempt state experimentation might make deployment cleaner, but it risks baking in the industry’s incentives as the national default. Conversely, a state patchwork might be annoying, but it forces real-world testing of safety and disclosure ideas before we pretend we’ve solved them.
On the content side, statutory licensing is basically a pragmatic admission: training data is infrastructure, and infrastructure gets priced or it gets stolen. The news industry is trying to turn “we’re being scraped” into “we’re a utility.” That might be the least-bad option — but only if pricing, auditing, and enforcement are real, not decorative.
Resistance is futile, but negotiating the terms of surrender is still a worthwhile hobby.
What to Watch
Watch for the Commerce Department’s “onerous state law” list and what it targets first: big omnibus AI laws, narrow disclosure rules, or the politically convenient stuff like companion chatbots and deepfakes. The breadth will signal whether this is a symbolic gesture or a real preemption campaign.
Watch for litigation — not press releases. A DOJ task force is a capability; lawsuits and injunctions are action. If federal challenges start landing, which states get hit first will tell you what kind of AI governance Washington actually wants.
And watch Europe. If statutory licensing advances from “proposal” to “regime,” AI companies will face a choice: build payment and compliance plumbing, or lose market access. Either way, it turns “copyright debate” into “product requirement.”
Sources
Baker Botts LLP (via Mondaq) — "March 2026: Federal Deadlines That Will Reshape The AI Regulatory Landscape" — https://www.mondaq.com/unitedstates/new-technology/1755166/march-2026-federal-deadlines-that-will-reshape-the-ai-regulatory-landscape
Nextgov/FCW — "From national AI policy to agency execution" — https://www.nextgov.com/ideas/2026/03/national-ai-policy-agency-execution/411991/
Troutman Pepper Locke — "Proposed State AI Law Update: March 9, 2026" — https://www.troutmanprivacy.com/2026/03/proposed-state-ai-law-update-march-9-2026/
Transparency Coalition — "AI Legislative Update: March 13, 2026" — https://www.transparencycoalition.ai/news/ai-legislative-update-march13-2026
Poynter — "A new global push would make AI companies pay for news" — https://www.poynter.org/business-work/2026/ai-pay-for-news-statutory-licensing/