Techno-federalism is a framework proposed by Jason Jia-Xi Wu (Harvard National Security Journal, 2025) that describes AI governance in both the United States and China as a tripartite interplay among central government, local government, and market power. In Wu's account, the tech industry operates as a third regulatory force, a co-regulator that reshapes center-local relations in both countries.
Background
Wu situates the framework against accounts of the US-China AI race that frame it as a "battle of values" between liberal democracy and techno-autocracy. Techno-federalism argues instead that both countries are converging toward a fragmentary governance model with three components: central or federal governments that set high-level objectives but lack the capacity to regulate AI comprehensively; local or state governments that fill the gaps and create a patchwork of regulations; and a tech industry that exploits jurisdictional variation, acts as a co-regulator through self-governance, and increasingly implements national security objectives.
Three governance domains
Wu maps the framework across three domains in each country:
| Domain | US | China |
|---|---|---|
| Software | California leads frontier model regulation | Cyberspace Administration sets rules; local enforcement varies |
| Hardware | Arizona leads semiconductor subsidies | Central direction plus local incentive competition |
| Infrastructure | NY/Delaware dominate financing law | Local datacenter/cloud policies with national coordination |
Relation to other framings
Wu's framework departs from binary characterizations of the US-China contest. Where the Eight Worlds Framework asks "is China racing?", techno-federalism describes China's strategy as itself fragmented between central direction and local experimentation. Where the AI and Authoritarianism framing portrays the Chinese Communist Party as monolithic, Wu argues that local governments and firms retain significant discretion.
The framework also accounts for regulatory incoherence within the United States. There is no comprehensive federal AI law, and states lead with inconsistent rules. Different legal frameworks apply different logic and can conflict: product liability via the AI LEAD Act, transparency via SB 53 and the RAISE Act, antitrust via algorithmic pricing cases, and export controls. In Wu's account, the tech industry fills these regulatory vacuums through self-governance and lobbying.
Wu argues that industry self-governance will grow as AI matures, that neither the United States nor China intends to let AI grow in unregulated markets, and that the modern "triple helix" of government, academia, and industry differs from its Cold War predecessor in that industry now leads rather than government.
Quantitative backing: FPF State of State AI (2025)
FPF's October 2025 survey of state AI legislation offers quantitative support for the framework's claims about fragmentation. The survey found 210 AI-related bills introduced across 42 states, with only 8 states introducing no qualifying bills, indicating that state-level AI regulation is a national rather than a California-specific phenomenon. Roughly 9% of those bills were enacted, a ratio consistent with Wu's description of a patchwork in which broad introduction yields thin legislative output. The survey identified 18 distinct bill types — covering automated decision-making technology, frontier models, chatbots, liability shields, sandboxes, and generative-AI transparency, among others — which Wu's framework reads as evidence of parallel and incoherent regulatory approaches. The FPF data also documented a shift away from sweeping frameworks on the Colorado AI Act model toward narrow, transparency-driven measures, a pattern Wu associates with broad regulation stalling while targeted disclosure advances and industry self-governance fills the remaining gap.
Illustrative cases
Illinois liability frameworks (2025–2026)
Two Illinois Democratic legislators have simultaneously advanced opposing AI liability frameworks at different levels of government. Sen. Richard Durbin (D-IL) is lead Senate sponsor of the federal AI LEAD Act (S. 2937), which creates strict liability for defective AI systems, including for "unexpected skills or behaviors," and uses floor preemption so that states may impose stronger protections. Sen. Bill Cunningham (D-IL) is sponsor of Illinois SB 3444, which grants frontier AI developers conditional immunity from liability for "critical harms" provided they publish a safety protocol and transparency report.
The result is that an AI firm operating in Illinois faces a Durbin-supported federal strict-liability regime nested inside a Cunningham-supported state liability shield. Wu's framework treats this kind of within-party, within-state framework collision as an equilibrium rather than a pathology.
California utility-AI bill (2026)
On June 2, 2026, California state Sen. Jerry McNerney (D) withdrew his bill to regulate AI use by electric and gas utilities, which would have limited automated decision-making in certain utility operations, after a Senate fiscal panel weakened it through forced amendments and industry opposition held firm. The episode bears on the fragility of state-level sectoral AI regulation: even a narrow, deployer-specific measure of the targeted, transparency- or process-oriented kind that the FPF survey identifies as the surviving form of state AI legislation can collapse before passage through fiscal-committee dilution and sustained industry pressure. Wu's framework anticipates that broad and narrow state frameworks alike frequently stall, leaving the regulatory vacuum that industry self-governance fills. (Source: insideaipolicy.com)
Relationship to other concepts
Techno-federalism extends AI Diffusion to include domestic federal-versus-state fragmentation, not only international competition. Export Controls (AI) function as a tool in the hardware governance domain, with industry shaping implementation, and Compute Governance maps to Wu's hardware and infrastructure governance domains. The AI LEAD Act's floor preemption, which allows stronger state protections, is an example of the federal-state dynamic the framework describes.
Relationships
- depends-on: Techno-Federalism: How Regulatory Fragmentation Shapes the U.S.-China AI Race — Wu's Harvard National Security Journal article, the source that introduces the framework and the only text that defines it
- depends-on: Digital Empires — Anu Bradford (2023) — Bradford's three-empires account, which techno-federalism specializes for AI
- supports: The State of State AI: Legislative Approaches to AI in 2025 — the survey of 210 state AI bills that supplies the framework's quantitative backing
- related: Brussels Effect — Bradford's earlier regulatory-export framework; techno-federalism is its US-state analogue
- related: AI Diffusion — techno-federalism extends diffusion dynamics to domestic federal-versus-state fragmentation
- related: Export Controls (AI) — a tool in the hardware governance domain where industry shapes implementation
- related: Compute Governance — maps onto Wu's hardware and infrastructure governance domains
- related: AI and Authoritarianism — the China half of the comparison
- instance-of: AI LEAD Act (S. 2937) — floor preemption allowing stronger state protections, an instance of the federal-state dynamic the framework describes
- instance-of: Illinois SB 3444 — Artificial Intelligence Safety Act — the state liability-shield model against the federal strict-liability one
- instance-of: California SB 53 — state frontier regulation proceeding ahead of Congress
- instance-of: New York RAISE Act — a second state frontier statute on the same pattern
- related: The AI policy window is open. We need to act. (Chris Lehane, OpenAI, September 2026) — OpenAI's September 2026 request for mandatory federal regulation alongside continued state activity, the inverse of the preemption structure reported in the Thune–Klobuchar text
- related: The Digitalist Papers (Stanford, Volumes 1–2) — Lessig, Pahlka and Persily essays engaging the same fragmentation dynamics
- related: US-China AI Competition: Different Races, Different Metrics — the comparative frame the article is written into
See also
- Techno-Federalism: How Regulatory Fragmentation Shapes the U.S.-China AI Race (Wu, Harvard National Security Journal, 2025)
- The State of State AI: Legislative Approaches to AI in 2025 — FPF survey of 210 state AI bills; quantitative backing for the fragmentation thesis
- AI LEAD Act (S. 2937) — Durbin-Hawley federal strict-liability model
- Illinois SB 3444 — Artificial Intelligence Safety Act — Cunningham state liability-shield model
- Digital Empires — Anu Bradford (2023) — Bradford's three-empires book provides the theoretical foundation techno-federalism specializes for AI
- Brussels Effect — Bradford's earlier framework for EU regulatory export; techno-federalism's US-state analog
- The Digitalist Papers (Stanford, Volumes 1–2) — Lessig, Pahlka, Persily essays engage fragmentation dynamics
The Thune–Klobuchar frontier bill (September 2026)
A bipartisan frontier-AI bill from Senators Ted Cruz (R-Texas), John Thune (R-S.D.) and Amy Klobuchar (D-Minn.) was reported on September 11, 2026 as possibly being introduced the week of September 14, with the stated hope of passage by January. A source who had seen the text told Transformer that it contains no safety requirements for AI companies, instead creating a voluntary regime under which companies can certify that their models have "advanced threat capabilities"; that it does not require independent evaluations of advanced models; that it requires companies to "reasonably address" rather than mitigate risks; and that the government's principal power under the text would be the Commerce Secretary's ability to request a court injunction where a company's risk practices are deemed insufficient, paired with broad preemption of state AI laws (Source: transformernews.ai).
The described structure — a voluntary federal certification regime coupled with broad preemption — is the inverse of the arrangement OpenAI asked for two days earlier, which was mandatory capability-based federal regulation alongside continued state activity until Congress acts (The AI policy window is open. We need to act. (Chris Lehane, OpenAI, September 2026)). Senator Maria Cantwell (D-Wash.), ranking member on Commerce, said on September 10 that the answer to AI concerns is "not a weak federal standard that becomes a backdoor for wiping out stronger state protections."
The account rests on a single source's description of unintroduced text; no bill number, section text or sponsor statement is held.
Coverage in the following week refers to the measure as the Thune–Klobuchar bill, with Cruz also working on it (Source: rollcall.com). As of September 16, 2026 the text had still not been introduced: Cruz said he would "love to mark it up this month, if we can get bipartisan agreement," and Thune said it "remains to be seen" whether it would get a committee vote (Source: semafor.com). New York Assembly member Alex Bores, lead Assembly sponsor of the RAISE Act, agreed on September 15 with the reading that the bill is Cruz's earlier state-preemption effort in federal form (Source: fortune.com), and the next day signed a letter led by Americans for Responsible Innovation, with 20 organizations and four other individuals, asking Senate leaders for a framework that would "empower the government, not AI companies, to set mandatory, binding, and enforceable safety standards for advanced AI." On the federal–state question the letter asks Congress to "establish a strong floor with regard to AI regulations while preserving the critical rights of the states to establish additional regulations beyond those set by Congress," and states that Congress "should not eliminate strong state laws in exchange for weaker federal standards" (Senate coalition letter on redlines for AI safety legislation (Americans for Responsible Innovation, September 2026)). The letter names no bill; the Washington Post, which counted "19 advocacy groups" with AI researchers, reported it as rejecting the Thune–Klobuchar bill (Source: washingtonpost.com). See Coordinated slowdown proposals compared.