El golpe federalista: la gran tecnología borra 50 leyes estatales

Key Takeaways

• Executive Order 14365, signed December 11, 2025, directs federal agencies to challenge, defund, and preempt state Artificial Intelligence (AI) laws by March 2026.
• The DOJ’s AI Litigation Task Force, formed January 9, 2026, will sue states like California, Colorado, and Texas whose AI regulations conflict with the new “minimally burdensome national standard.”
• The FTC’s legal authority to preempt state law is extremely limited, requiring years of formal rulemaking under the Magnuson-Moss Act. The Executive Order (EO) is a political threat, not an immediate legal override.
• The real beneficiaries are not startups. Cross-state compliance burdens favor incumbents. A weak federal “floor” replacing strong state “ceilings” is the goal of $125 million in lobbying by Andreessen Horowitz (a16z) and allies.
• The historical parallel is the 1996 Telecommunications Act, which promised competition through federal preemption but delivered consolidation into three carriers. AI is getting the same treatment.

The 90-Day Countdown

On December 11, 2025, President Trump signed Executive Order 14365, titled “Ensuring a National Policy Framework for Artificial Intelligence.” The title sounds anodyne, the language bureaucratic. But the mechanism is anything but.

By March 11, 2026, the Secretary of Commerce must publish a list of state AI laws deemed “onerous.” The Federal Trade Commission (FTC) must issue a policy statement claiming that Section 5 of the FTC Act preempts those state laws. And the Attorney General’s new “AI Litigation Task Force,” formed on January 9, 2026, will begin filing lawsuits against states whose regulations the administration finds inconvenient.

This is not a policy debate. It is a constitutional coup, conducted in the dry language of administrative procedure. And the beneficiaries are not the American people, nor the scrappy startups that Silicon Valley claims to champion. The beneficiaries are the trillion-dollar incumbents who wrote the lobbying checks.

The State Laws Under Attack

Before understanding the preemption, you must understand what is being preempted. As of February 2026, a wave of state AI laws has come into effect, representing the most significant attempt to regulate algorithmic systems since the European Union’s General Data Protection Regulation (GDPR).

California: SB 53 (The Transparency in Frontier Artificial Intelligence Act)

Effective January 1, 2026, California’s SB 53 is the first state-level AI safety law in the country. It applies to any developer who trains a model using more than $10^{26}$ floating-point operations (FLOPs), a threshold that targets only the largest frontier labs like OpenAI, Anthropic, and Google DeepMind.

Key Requirements:

• Publish a written framework explaining how national and international safety standards are applied during model development.
• Mandatory reporting of “critical safety incidents” to the California Office of Emergency Services.
• Public transparency reports every three months detailing model releases and internal use.
• Robust whistleblower protections for employees raising AI safety concerns.

Enforcement: Civil penalties up to $1 million per violation, enforceable by the California Attorney General.

Colorado: The Anti-Discrimination in AI Law

Scheduled for June 2026, Colorado’s law requires employers to conduct risk assessments on high-risk AI systems. It mandates transparency notices to candidates and employees when AI influences hiring or firing decisions. The standard is “reasonable care” to prevent algorithmic discrimination.

New York: The RAISE Act

Signed December 16, 2025, and effective January 1, 2027, New York’s Responsible AI Safety and Education Act mirrors California’s approach, applying to frontier models with high computational thresholds and mandating safety protocols.

Texas: HB 149 and SB 815

Texas has enacted laws requiring AI disclosures and establishing an AI advisory council, with specific rules for AI in health insurance decisions.

Florida: SB 482

Even as EO 14365 was signed, Florida’s legislature introduced SB 482 to regulate AI chatbots interacting with minors. Governor DeSantis has publicly asserted the state’s Tenth Amendment right to proceed.

These state laws share a common thread: they make AI developers accountable for the systems they deploy. They require transparency, safety reporting, and anti-discrimination measures. They treat algorithmic decision-making as a matter of public concern, not just private commerce.

And that is precisely why Big Tech wants them destroyed.

The Preemption Playbook

EO 14365 does not outright ban state AI laws. Executive Orders lack that power; only Congress can preempt state law under the Constitution’s legislative powers. What the EO accomplishes is far more insidious: it weaponizes the federal bureaucracy to make state enforcement economically and legally impossible.

Mechanism 1: The Funding Threat

The EO directs the Secretary of Commerce to issue a policy within 90 days stating that any state imposing certain AI requirements may be ineligible for broadband funding under the Broadband Equity Access and Deployment (BEAD) Program. BEAD represents billions of dollars in federal infrastructure money. For cash-strapped state governments, losing access to that funding is existential.

This is not a legal argument. It is coercion. The federal government is threatening to defund states that regulate AI in ways Big Tech dislikes.

Mechanism 2: The FTC Policy Statement

The EO tasks the FTC with issuing a policy statement claiming that Section 5 of the FTC Act (which prohibits “unfair or deceptive acts or practices”) preempts state laws requiring AI developers to alter model outputs in ways that produce “less truthful” results.

The logic is tortured: Colorado’s anti-discrimination law, for instance, requires AI systems to avoid biased outputs. The administration claims this forces developers to produce “false results” by mandating ideological bias corrections. Ergo, the FTC’s ban on deception preempts the state law.

This argument fails basic legal scrutiny. As legal scholars at Tech Policy Press have noted, Section 5 lacks explicit preemption language. Courts apply a presumption against preemption absent clear congressional intent. The FTC would need to conduct formal rulemaking under the Administrative Procedure Act (APA) and the Magnuson-Moss Act, a process requiring public notice, comment periods, regulatory impact analyses, and potentially years of litigation.

EO 14365 is not a legal override. It is a political threat designed to chill state enforcement while the lobbying machine grinds toward a congressional solution.

Mechanism 3: The DOJ AI Litigation Task Force

On January 9, 2026, Attorney General Pam Bondi formed the AI Litigation Task Force, as directed by EO 14365. Its mandate: challenge state AI laws on federal preemption and Dormant Commerce Clause grounds.

The Dormant Commerce Clause argument is straightforward: state laws that regulate beyond their borders or impose undue burdens on interstate commerce are unconstitutional. Since AI models are trained and deployed nationally, any state-specific requirement arguably affects interstate commerce.

This is the same argument the telecom industry used in 1996. It worked then. It is being deployed again now.

The 1996 Playbook: A History Lesson

The Telecommunications Act of 1996 was sold as a deregulatory triumph. By preempting the “patchwork” of state telecom rules, Congress would spark competition, lower prices, and bring innovation to every corner of America.

Thirty years later, the results are in. The local telecom market is dominated by three carriers: AT&T, Verizon, and T-Mobile. Local competition is dead. And the FCC has progressively expanded preemption far beyond the original statutory language, using “deemed granted” shot clocks and administrative orders to override municipal zoning and force 5G tower installations over local objections.

A 2022 report titled “Feeling the Squeeze” found that FCC preemption negatively impacted municipal finances, created public safety challenges, and interfered with digital equity efforts. Twenty-nine cities have formally objected to the FCC’s ongoing “Build America” agenda, which seeks to eliminate remaining state and local regulatory authority.

The 1996 model did not deliver competition. It delivered consolidation. The weak federal “floor” replaced stricter state “ceilings,” leaving incumbents as the clear winners.

AI is getting the same treatment.

Follow the Money: The Lobbying Machine

Behind every regulatory rollback is a lobbying operation. The AI preemption push is no exception.

Leading the Future: The $125 Million Super PAC

Launched in August 2025 with an initial $100 million announcement, Leading the Future is the political action committee (PAC) at the center of the anti-regulation campaign. As of January 2026, it has raised over $125 million to influence the 2026 midterms.

Major donors include:

• Andreessen Horowitz (a16z)
• OpenAI co-founder Greg Brockman
• 8VC’s Joe Lonsdale
• SV Angel’s Ron Conway
• Perplexity AI

The PAC’s stated goal: elect federal candidates who support a “responsible national framework” for AI—code for weak federal standards that preempt stricter state rules. Affiliate organizations like Build American AI run ads attacking state legislation. Industry-funded campaigns contributed to the veto or dilution of bills like New York’s RAISE Act and California’s LEAD for Kids Act.

David Sacks: The Crypto-AI Czar

David Sacks, co-host of the “All-In” podcast and general partner at Craft Ventures, was appointed as Trump’s White House AI and Crypto Advisor. Sacks is the architect of EO 14365.

Craft Ventures has extensive investments in AI and cryptocurrency startups. Sacks personally benefits from policies that reduce regulatory friction for those investments. Senator Elissa Slotkin (D-MI) has called for an investigation into potential conflicts of interest.

The revolving door between Silicon Valley venture capital and the Trump administration is not subtle. It is the policy.

The AI Infrastructure Coalition

A16z has co-founded the AI Infrastructure Coalition with hyperscalers Google, Meta, Microsoft, and (notably) ExxonMobil. The coalition lobbies against data center scrutiny and pushes for accelerated energy permitting to support AI compute expansion. The interests are aligned: more AI, less oversight, faster permitting.

The Constitutional Clash

EO 14365 sets up a constitutional confrontation between federal supremacy and state police powers.

The Federal Position: Supremacy and Commerce

The Supremacy Clause of the Constitution establishes federal law as the “supreme Law of the Land.” When federal and state laws conflict, federal law prevails. The administration’s argument is simple: AI is inherently interstate commerce, and a patchwork of 50 state laws imposes unconstitutional burdens.

The Dormant Commerce Clause, a judicial doctrine rooted in the Constitution’s commerce provisions, prohibits states from passing laws that discriminate against or unduly burden interstate commerce. Since AI models are trained on national datasets and deployed globally, any state-specific rule arguably interferes.

The State Position: Tenth Amendment and Police Powers

The Tenth Amendment reserves to the states all powers not delegated to the federal government. Historically, this includes “police powers” over health, safety, and consumer protection.

Florida’s Governor DeSantis has explicitly invoked the Tenth Amendment to justify continuing with SB 482 despite EO 14365. California and New York are likely to mount similar defenses. The argument: regulating algorithmic discrimination is a consumer protection function, not an interference with commerce.

The Legal Reality: No Automatic Preemption

Here is what EO 14365 does not do: it does not automatically invalidate a single state law. Executive Orders cannot override state legislation. Only Congress can preempt state law, and only if the preemption is clear and explicit.

The FTC’s Section 5 authority is not a preemption statute. Courts have consistently required formal rulemaking, not policy statements, to establish preemption. That rulemaking would take years.

What EO 14365 does is create legal uncertainty. State AGs may hesitate to enforce. Business compliance departments may adopt a “wait and see” posture. The chilling effect is the point.

Who Actually Benefits?

The industry narrative is that federal preemption helps “innovation” by freeing startups from compliance burdens. This is marketing, not reality.

The Compliance Cost Asymmetry

Cross-state compliance is expensive. A company deploying AI in all 50 states must theoretically track 50 different regulatory regimes. But here is the asymmetry: who can afford 50 compliance lawyers?

The answer is Google, Meta, OpenAI, and Microsoft. The companies with trillion-dollar valuations can absorb compliance costs as a rounding error. Startups cannot.

A “patchwork” of strong state laws actually helps smaller players who can use ethical compliance as a competitive differentiator. “Built to exceed California standards” is a marketing advantage when California is the world’s fifth-largest economy.

Federal preemption eliminates that advantage. It replaces strong state ceilings with a weak federal floor—the same floor the largest incumbents helped write. The barrier to entry becomes political access, not product quality.

The GDPR Paradox

Consider the precedent. When the EU enacted GDPR, American companies complained bitterly about compliance costs. Then a funny thing happened: GDPR became the de facto global privacy standard. American firms comply with GDPR for their European operations, and many extend those protections globally.

If EO 14365 succeeds in preempting state AI laws, the vacuum will not remain empty. The EU’s AI Act, with fines up to €35 million or 7% of global turnover, will become the binding constraint for any company operating internationally. American AI governance will be written in Brussels, not Washington.

Federal preemption does not eliminate regulation. It exports regulatory authority to jurisdictions that actually apply it.

The March Deadline and What Comes Next

The 90-day clock expires on March 11, 2026. By that date:

1. Commerce Department must publish a list of “onerous” state AI laws and issue a policy on BEAD funding ineligibility.
2. FTC must issue a policy statement on Section 5 preemption.
3. DOJ AI Litigation Task Force will begin challenging state laws in court.

Congressional allies are pushing to codify EO 14365 into statute. Rep. Jay Obernolte (R-CA), chair of the House AI Task Force, is developing a “hub and spoke” federal framework that would pair minimal national standards with sector-specific rules while preempting conflicting state laws. As of February 5, 2026, the bill has not been formally introduced. House GOP leadership has sidelined Obernolte from high-level negotiations.

The coalition supporting preemption is well-funded. But the constitutional barriers are real. Courts will not rubber-stamp executive overreach. State AGs in California, New York, and Florida are already preparing challenges.

The Stakes

This is not an abstract policy debate. It is a test case for American federalism in the age of technology.

The question is simple: Who gets to regulate the systems that decide who gets a job, a loan, a medical diagnosis, or a seat on an airplane?

If the answer is “nobody,” because federal preemption has erased state authority and Congress has failed to act, then accountability disappears. The trillion-dollar platforms operate in a governance vacuum, constrained only by market competition—which, as the telecom sector demonstrates, tends toward consolidation, not choice.

If the answer is “states,” then California, Colorado, and New York become the laboratories of democracy the founders envisioned. Their experiments can fail or succeed on their own merits, providing empirical evidence for what works.

EO 14365 is an attempt to foreclose that experiment before the results are in. The $125 million lobbying campaign is not about innovation. It is about control: ensuring the rules of the AI future are written by the companies that stand to profit, not the citizens who stand to be governed.

The Resistance

The good news: the preemption campaign is not going unchallenged.

State AGs from California, New York, Illinois, and potentially Florida are preparing legal responses. Civil liberties organizations are documenting the lobbying connections. Senate Democrats like Elissa Slotkin, Amy Klobuchar, and Cory Booker have demanded ethics investigations.

And the constitutional structure itself is a check. Executive Orders cannot override state law. The FTC lacks the statutory authority for broad preemption. Congressional action requires bipartisan support, and the GOP is divided between innovation-friendly moderates and Tenth Amendment purists.

The outcome is not predetermined. But the sides are clear: trillion-dollar incumbents versus democratic accountability.

The next 90 days will determine which side wins.