Law and The Machine

The "Light Touch" Collision: Federal Preemption, Copyright, and the Trump AI Framework

April 03, 202618:42Law and The Machine

This episode explores the escalating conflict over AI governance in American federalism, revealing how the federal government is using legal challenges and financial incentives to impose a deregulatory "light touch" AI framework on states. Listeners will learn about the Department of Justice's "AI Litigation Task Force," the contradictory legislative efforts, and the significant conflicts of interest influencing federal AI policy.

Key Takeaways

Detailed Report

A significant 'civil war' has erupted over AI governance in the United States, pitting the White House against Congress and individual states, with profound implications for federalism, intellectual property, and ethical AI development.

The Federal 'Light Touch' Framework

The Trump administration has introduced a 'National Policy Framework for Artificial Intelligence,' advocating a 'light touch' approach. This framework, co-authored by White House 'AI & Crypto Czar' David Sacks, who maintains a venture capital fund with AI investments, demands federal preemption of state AI laws, broad fair-use protections for training data, and a Section 230-style liability shield for model developers.

This strategy stems from Executive Order 14365, signed last December, which aims to dismantle a 'patchwork' of 50 state regulations perceived as stifling innovation. To enforce this, the Department of Justice (DOJ) launched an 'AI Litigation Task Force' on January 9th. Its mandate is to actively challenge state AI laws in court, arguing they 'unconstitutionally regulate interstate commerce, are preempted by existing Federal regulations, or are otherwise unlawful.' This approach marks a significant departure from traditional conservative 'states' rights' principles, with the federal government using its power to shield tech companies from compliance costs.

Clashing Views on Intellectual Property

One of the most explosive collisions is over intellectual property. The White House's March 20th framework explicitly declares that 'training AI models on copyrighted material does not violate copyright law,' framing it as Fair Use. This position is a major win for the tech industry, effectively blessing the mass scraping of web data without licensing or compensation.

However, just two days prior, Senator Marsha Blackburn introduced her 'TRUMP AMERICA AI Act.' Despite its name, this 291-page bill directly contradicts the White House, establishing national rules to protect creators from unauthorized AI training and AI-generated knockoffs. It incorporates provisions from the NO FAKES Act and the Kids Online Safety Act. If passed, Blackburn's bill would fundamentally reshape the AI landscape, making the cost of licensing training data astronomical and creating an 'unnavigable legal minefield' for developers.

The Liability Shield Debate

The White House Framework also includes a directive targeting state laws that 'penalize AI developers for a third party's unlawful conduct involving their models,' effectively proposing a Section 230-style liability shield for generative AI. This means if a user creates harmful content with an AI model, the model creator would not be held liable by states.

In stark contrast, Senator Blackburn's bill seeks to completely repeal Section 230 and explicitly establishes a 'duty of care' for AI developers. It would allow state attorneys general and private actors to sue AI developers for physical, financial, and reputational harms, creating a significant accountability gap between the executive and legislative branches.

Economic Coercion: Weaponizing Broadband Funds

Beyond legal frameworks, the federal government is also employing economic leverage. The December Executive Order directed the National Telecommunications and Information Administration (NTIA) to condition the disbursement of leftover Broadband Equity, Access, and Deployment (BEAD) funds on states repealing 'onerous' AI laws. These funds, originally earmarked for rural broadband and digital equity, amount to a $21 billion pool.

This threat has already had real-world impact. On April 1st, members of the Louisiana state legislature withdrew a package of bipartisan AI regulatory bills after administration officials explicitly threatened to withhold the state's BEAD funding. This demonstrates a clear instance of the federal government holding critical infrastructure funding hostage to influence state-level AI policy.

The Ethics Penalty: Government as Customer

The Pentagon's actions illustrate how the government, as a major AI customer, can shape the industry's ethical landscape. In July 2025, Anthropic, an AI safety-focused lab, signed a $200 million contract with the Pentagon for its Claude model, with strict ethical red lines against mass domestic surveillance and integration into lethal autonomous weapons.

However, when the Department of Defense later demanded unrestricted 'all lawful use' access, Anthropic refused to remove its guardrails. In retaliation, on February 27, 2026, Defense Secretary Pete Hegseth designated Anthropic a 'supply chain risk to national security,' ordering all federal agencies to cease commercial activity with the company. Internal memos revealed this designation was based on Anthropic's 'increasingly hostile manner' in refusing the terms, not actual cybersecurity vulnerabilities.

Less than 24 hours later, OpenAI signed a contract with the DOD for the exact same classified networks. While OpenAI later added an addendum claiming similar red lines, experts noted crucial loopholes, such as prohibiting 'deliberate' but not 'incidental' surveillance. Anthropic was punished for upholding its principles, while OpenAI, with more pliable terms, secured the lucrative contract. This sends a chilling message to AI companies about the value of ethical guardrails in government partnerships.

Broader Implications

The current landscape presents a messy, hybrid reality for the tech industry. The executive-led 'light touch' approach, enforced through litigation and financial coercion, appears designed to protect the business models of incumbent tech giants. Meanwhile, the stark divergence between the White House and Congress on critical issues like copyright and liability creates an unstable legal environment for innovation.

This raises fundamental questions about the future of democratic oversight in a rapidly evolving technological landscape, particularly when the federal government uses strong-arm tactics to dictate AI policy and when corporate ethics become a 'disqualifying liability' in the pursuit of government contracts.

Show Notes

Works Referenced

  • National Policy Framework for Artificial Intelligence: A "light touch" approach to AI regulation proposed by the Trump administration on March 20th, advocating for federal preemption, broad fair-use protections, and liability shields.
  • TRUMP AMERICA AI Act: A 291-page bill introduced by Senator Marsha Blackburn that contradicts the White House framework on copyright, liability, and state preemption, aiming to protect creators and establish developer duties.
  • DOJ's AI Litigation Task Force: Launched by Attorney General Pam Bondi on January 9th, this task force actively challenges state AI laws in court, arguing they unconstitutionally regulate interstate commerce or are preempted by federal regulations.
  • Executive Order 14365, "Removing Barriers to American Leadership in Artificial Intelligence": A December executive order that laid the groundwork for the federal government's deregulatory AI strategy and directed the conditioning of BEAD funds.
  • BEAD (Broadband Equity, Access, and Deployment) funds: Federal funds initially earmarked for rural broadband, now being used by the administration to pressure states into complying with its AI agenda by threatening to withhold money.
  • Section 230: A provision of the Communications Decency Act that protects online platforms from liability for third-party content, which the White House framework seeks to extend to AI models.
  • Anthropic: An AI safety-focused lab that signed a contract with the Pentagon but was later designated a "supply chain risk" for upholding ethical guardrails against mass surveillance and autonomous weapons.
  • OpenAI: A rival AI lab that quickly replaced Anthropic on the Pentagon's GenAI.mil network, signing a contract with similar-sounding but more pliable ethical terms.
  • Michael Kratsios: Former OSTP Director and co-author of the "National Policy Framework for Artificial Intelligence."
  • David Sacks: White House "AI & Crypto Czar" and co-author of the AI framework, who operates under an ethics waiver allowing him to maintain his venture capital fund, Craft Ventures, with investments in AI.
  • Attorney General Pam Bondi: Formally launched the DOJ's AI Litigation Task Force.
  • Senator Marsha Blackburn: Introduced the "TRUMP AMERICA AI Act" and is a key figure in the "Copyright Civil War."
  • Baker Botts: Legal analysts who noted the unprecedented nature of the DOJ's preemption strategy through litigation rather than Congressional action.
  • Kathleen Clark: An ethics expert who described special government employee waivers as "a presidential pardon in advance" for conflict-of-interest statutes.
  • NO FAKES Act: Legislation incorporated into Senator Blackburn's bill, aimed at protecting creators from AI-generated knockoffs.
  • Kids Online Safety Act: Legislation also incorporated into Senator Blackburn's bill, addressing online safety for children.
  • Cato Institute: A think tank that has been highly critical of the Blackburn bill, warning it would create an "AI-audit industrial complex."
  • Latham & Watkins: A law firm that issued a memo noting Blackburn’s bill would "rewrite copyright liability."
  • Bill Baer: A Visiting Fellow at Brookings who criticized the "empty national AI policy framework" for ignoring the concentration of AI decision-making power.
  • National Telecommunications and Information Administration (NTIA): The agency directed to condition the disbursement of BEAD funds on states repealing "onerous" AI laws.
  • FTC (Federal Trade Commission): A federal regulator that, in February, reopened and set aside a 2024 consent order against Rytr, signaling a shift in federal oversight.
  • Rytr: A generative AI company whose consent order was withdrawn by the FTC, aligning with the White House's deregulatory mandate.
  • Defense Secretary Pete Hegseth: Took the unprecedented step of designating Anthropic a "supply chain risk to national security" for upholding its ethical guardrails.
  • Sam Altman: CEO of OpenAI, who called the company's initial rollout with the DOD "opportunistic and sloppy" amidst public backlash.
  • Jake Laperruque: A tech policy expert who pointed out the crucial loopholes in OpenAI's contract with the DOD regarding surveillance.
  • Craft Ventures: David Sacks' venture capital fund, which has significant investments in the AI sector.

Glossary

  • Federal preemption: The legal principle where federal law takes precedence over state law when there is a conflict or the federal government intends to occupy a field exclusively.
  • Fair Use: A legal doctrine in copyright law that allows limited use of copyrighted material without permission from the rights holder, such as for commentary, criticism, news reporting, or research.
  • Section 230: A provision of the Communications Decency Act of 1996 that states "No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider," largely shielding online platforms from liability for third-party content.
  • Algorithmic transparency: The ability to understand how an algorithm works, what data it uses, and how it arrives at its decisions, often sought by regulators to ensure fairness and accountability.
  • Regulatory capture: A form of government failure where a regulatory agency, created to act in the public interest, instead advances the commercial or political concerns of special interest groups that dominate the industry or sector it is supposed to regulate.
  • LLM (Large Language Model): An artificial intelligence program trained on vast amounts of text data to understand, generate, and respond to human language in a coherent and contextually relevant way.
  • Deepfakes: Synthetic media, typically videos or images, that have been digitally manipulated using AI to replace one person's likeness with another's, often for malicious purposes.
  • BEAD (Broadband Equity, Access, and Deployment): A federal program established by the Infrastructure Investment and Jobs Act to fund high-speed internet infrastructure deployment and adoption programs in underserved communities across the United States.
  • Lethal autonomous weapons systems: Weapons systems that can select and engage targets without human intervention, raising significant ethical and legal concerns.
  • GenAI.mil: A classified military network mentioned in the episode where AI models are deployed for defense purposes.
  • Special government employee (SGE): An individual who performs temporary duties for the government, often allowed to retain private sector interests, sometimes with ethics waivers, due to their specialized expertise.
  • Frontier models: The most advanced and capable artificial intelligence models currently available, often characterized by their large scale and broad range of applications.

Full Transcript

HostYou know, we've talked a lot about the theoretical battles over AI governance – who should regulate, what the rules should be. But there's a particular executive action that just dropped that completely flips the script on what we thought we knew about American federalism.
ExpertIt's more than theoretical now, isn't it? We're seeing a full-blown civil war erupt between the White House, Congress, and literally all fifty states. And what's wild is the *tools* being deployed by the federal government to win this war.
HostExactly. Forget the usual states' rights arguments. We're talking about the Department of Justice, armed with an "AI Litigation Task Force," actively suing states to prevent them from regulating technology. This isn't just a policy disagreement; it's a legal blitzkrieg.
ExpertAnd it’s not just legal action. There’s money on the table, too. Taxpayer money, initially earmarked for rural broadband, now being weaponized to strong-arm states into compliance with a very specific, deregulatory AI agenda.
HostIt’s a fascinating, and frankly, alarming, moment. On one side, you have the Trump administration pushing what they call a "National Policy Framework for Artificial Intelligence," a "light touch" approach drafted by folks like OSTP Director Michael Kratsios and Special Advisor David Sacks. It demands federal preemption of state AI laws, broad fair-use protections for training data, and a Section 230-style liability shield for model developers.
ExpertBut on the other side, and almost simultaneously, Senator Marsha Blackburn introduced her "TRUMP AMERICA AI Act." Despite the name, this bill is anything but light touch. It's a 291-page regulatory behemoth that fundamentally contradicts the White House on almost every front: copyright, liability, and state preemption. It’s like they’re living in different universes.
HostIt really is. Let's unpack this "light touch" playbook first, because it comes with some incredibly heavy-handed enforcement. The current federal strategy, according to the research, flows from an Executive Order signed last December, EO 14365, "Removing Barriers to American Leadership in Artificial Intelligence." The stated goal is to dismantle what the tech lobby calls a "patchwork" of 50 state regulations that stifle innovation.
ExpertAnd that culminated in the March 20th framework, which outlines seven pillars: child safety, community protections, intellectual property, free speech, innovation, workforce development, and, the one we’re really focused on, federal preemption of state AI laws. It's only four pages long, non-binding, but it’s a clear statement of intent.
HostAnd the enforcement arm behind this non-binding framework is what truly caught my attention. On January 9th, Attorney General Pam Bondi formally launched the DOJ's "AI Litigation Task Force." Its mandate is to explicitly challenge state AI laws in court, arguing they "unconstitutionally regulate interstate commerce, are preempted by existing Federal regulations, or are otherwise unlawful."
ExpertAs an engineer, you hear "patchwork of regulations" and you can understand the industry's desire for uniformity. It makes development and deployment easier. But the way they're going about it…
HostThat's where the legal brain kicks in. This completely flips the traditional conservative "states' rights" narrative on its head. The federal government, historically the champion of less federal overreach, is now actively using taxpayer dollars to sue states to prevent them from regulating technology. Legal analysts at Baker Botts pointed out that while past preemption regimes, like the Airline Deregulation Act, were products of *Congressional action*, this Task Force is attempting to achieve regulatory uniformity through "litigation and Executive coordination alone." That's a massive power grab.
ExpertIt's like the federal government is saying, "We want a free market, but only if we define what 'free' means, and we'll sue you if you disagree."
HostAnd who's behind this "light touch" framework? It was co-authored by David Sacks, who serves as the White House "AI & Crypto Czar." But here’s the kicker: Sacks operates under a "special government employee" designation, which grants him a controversial ethics waiver. This allows him to maintain his venture capital fund, Craft Ventures, which has significant investments in the AI sector.
ExpertWait, so the person helping draft the rules for AI also has direct financial stakes in AI companies? That doesn't just sound like a conflict of interest, it *is* a conflict of interest.
HostAccording to ethics expert Kathleen Clark, these waivers act "like a presidential pardon in advance" for conflict-of-interest statutes. And the payoff is clear: Sacks has actively pushed for the preemption of state laws that mandate algorithmic transparency – laws that would directly increase compliance costs for his portfolio companies. It’s regulatory capture, plain and simple, but with a presidential waiver.
ExpertIt's a convenient arrangement, to say the least. From a technical perspective, transparency mandates *can* be challenging to implement, especially for proprietary models. But that's usually part of the cost of doing business, not something you get a federal waiver to avoid.
HostAnd this brings us to another massive collision, perhaps the most explosive one happening right now: intellectual property. The White House, in its March 20th framework, explicitly declares that "training AI models on copyrighted material does not violate copyright law," framing it as Fair Use.
ExpertThat's a huge deal for the tech industry. For model developers, it means they can effectively ingest petabytes of proprietary data without having to worry about licensing or compensation. It's the tech lobby's dream scenario, essentially blessing the current practice of scraping the web en masse.
HostBut then, just two days *before* the White House framework came out, Senator Marsha Blackburn released her discussion draft for the "TRUMP AMERICA AI Act." And despite the flattering title, it’s a direct counter-punch to the White House on IP. Her bill explicitly establishes national rules to protect creators from unauthorized AI training and AI-generated knockoffs. It even incorporates the NO FAKES Act and the Kids Online Safety Act.
ExpertIf Blackburn's bill passes, it would fundamentally reshape the AI landscape. The sheer cost of licensing training data would be astronomical. It would likely bankrupt many open-source projects and mid-tier AI labs. Only the trillion-dollar incumbents, companies like Meta or Google, who already have massive, walled-garden data ecosystems, could afford to train frontier models. The Cato Institute, for example, has been highly critical of the Blackburn bill, warning it would create an "AI-audit industrial complex" and force developers to "serially delay deployment."
HostSo, on one hand, the executive branch is telling AI developers to scrape the web freely, and on the other, the Senate is drafting legislation that would establish sweeping liability for those exact actions. As a March 26th memo from Latham & Watkins put it, Blackburn’s bill would "rewrite copyright liability." It creates an unnavigable legal minefield for developers.
ExpertAbsolutely. Imagine being a startup right now. Do you invest millions in training a model on publicly available data, hoping the White House's "fair use" stance holds, or do you assume Blackburn's bill is the future and start trying to license everything, which might not even be possible? It's a completely unstable legal environment.
HostAnd it gets even more ironic. Tucked into the White House Framework is a directive targeting state laws for preemption if they "penalize AI developers for a third party's unlawful conduct involving their models."
ExpertLet me guess: they're trying to create a Section 230-style liability shield for generative AI.
HostPrecisely. If a user utilizes an open-source LLM to generate a mass phishing campaign, or uses an image generator to create non-consensual deepfakes, the White House position is that the state cannot hold the model creator liable.
ExpertThe sheer irony of the Trump administration, which has been one of the most vocal opponents of Section 230 for social media platforms, now proposing a virtually identical liability shield for generative AI models, is just… it's something else.
HostIt's an accountability gap, pure and simple. Meanwhile, the Blackburn bill seeks to completely *repeal* Section 230 and explicitly establishes a "duty of care" for AI developers. It would allow state attorneys general and private actors to sue AI developers for physical, financial, and reputational harms. The divergence couldn't be starker.
ExpertIt really highlights what Bill Baer, a Visiting Fellow at Brookings, called the "empty national AI policy framework." He wrote that it's a "laundry list" of symptoms that completely ignores the root cause: the concentration of AI decision-making power. As he puts it, "Power does not regulate itself."
HostAnd it's not just the DOJ and legal frameworks being used as leverage. The federal government is also weaponizing taxpayer dollars in a truly cynical way. The December Executive Order directed the National Telecommunications and Information Administration, or NTIA, to condition the disbursement of leftover BEAD — Broadband Equity, Access, and Deployment — funds on states repealing "onerous" AI laws.
ExpertSo, these are funds originally earmarked by Congress for rural broadband and digital equity, right? We're talking about connecting underserved communities, critical infrastructure.
HostThat's right. And due to a controversial restructuring of the BEAD program, there's currently a $21 billion pool of "non-deployment" funds sitting in limbo. The administration is now threatening to withhold this money from any state that attempts to regulate AI safety or algorithmic discrimination.
ExpertThis isn't theoretical. There's a real-world impact already.
HostThere is. On April 1st, members of the Louisiana state legislature officially withdrew a package of bipartisan AI regulatory bills. State legislators reported that administration officials explicitly threatened to withhold the state's BEAD funding if the bills advanced. Democratic State Senator Jay Luneau stated plainly: "We stand to lose a lot of money in the state of Louisiana that would help a lot of people if we go forward with this bill."
ExpertSo, the federal government is holding internet access for rural communities hostage to protect the profit margins of Silicon Valley AI labs from state-level consumer protection laws. That's a pretty stark example of power at play.
HostIt's transactional, raw, and frankly, astonishing. It leaves the tech industry in a messy, hybrid reality. Neither the White House Framework nor the Blackburn bill are binding law yet. But the market, as the research shows, is reacting as if the "light touch" era is already permanent.
ExpertYou see AI labs accelerating model training and data center expansion without fear of federal reprisal, banking on the DOJ's Task Force to fight their state-level legal battles for them. And even federal regulators are falling in line. The FTC, in February, reopened and set aside its 2024 consent order against Rytr, a generative AI company. That withdrawal of a Biden-era enforcement action signals that federal regulators are actively shedding their oversight responsibilities to comply with the White House's deregulatory mandate. It's a clear signal from the top.
HostAnd speaking of signals, this brings us to The Conflict Docket. We've got a case here that perfectly illustrates how the government, acting as a major AI customer, can shape the entire ethical landscape of the industry.
ExpertThis is where we really see the lines blur between regulator, contractor, and lobbyist.
HostAbsolutely. The case involves the Pentagon, Anthropic, and eventually, OpenAI. Back in July of 2025, Anthropic, an AI safety-focused lab, signed a $200 million contract with the Pentagon to deploy its frontier model, Claude, on the military's classified "GenAI.mil" network.
ExpertAnthropic has been pretty vocal about its ethical red lines, right? They built them into their acceptable use policy.
HostExactly. Their policy contained strict red lines: Claude could not be used for mass domestic surveillance of American citizens, nor could it be integrated into lethal autonomous weapons systems that lacked a human decision-maker. Fast forward to early 2026, and the Department of Defense demands unrestricted, "all lawful use" access to the technology. Anthropic, sticking to its principles, refused to remove its ethical guardrails.
ExpertAnd what happened?
HostIn retaliation, on February 27, 2026, Defense Secretary Pete Hegseth took an unprecedented step: he officially designated Anthropic a "supply chain risk to national security." He then ordered all federal agencies and military contractors to cease commercial activity with the company.
ExpertA "supply chain risk"... for upholding their ethical guidelines? That's quite a classification.
HostInternal Pentagon memos later revealed that this designation was not based on actual cybersecurity vulnerabilities, but rather on Anthropic's "increasingly hostile manner" in refusing the surveillance and weapons terms. And here’s where it gets even more pointed: Less than 24 hours after Anthropic was blacklisted, rival AI lab OpenAI announced it had signed a contract with the DOD to deploy its models on the *exact same* classified networks.
ExpertWow. That's... quick.
HostThere was massive public backlash, including a 295% overnight spike in ChatGPT app uninstalls and the resignation of OpenAI's hardware lead. OpenAI CEO Sam Altman called their initial rollout "opportunistic and sloppy." Then, on March 2nd, OpenAI released an addendum to their contract, claiming they had successfully negotiated the *exact same* red lines that got Anthropic blacklisted: no mass domestic surveillance, no autonomous weapons, and no high-stakes automated decisions.
ExpertSo, they got the same terms. Why was Anthropic blacklisted then?
HostThat’s the critical question, and tech policy experts like Jake Laperruque point to the fine print. OpenAI's contract prohibits "deliberate" and "intentional" surveillance of Americans, but it leaves massive loopholes for "incidental collection" within large datasets purchased from commercial data brokers. Furthermore, the only enforcement mechanism OpenAI has is the right to walk away from the contract if the military violates the terms. That's a right Anthropic exercised, resulting in their virtual banishment from government contracting.
ExpertSo, Anthropic held firm and was punished. OpenAI signed a contract with similar-sounding terms, but with crucial loopholes, and immediately got the business. It paints a very clear picture of what kind of ethical posture the government values in its AI partners.
HostIt does. When the Defense Department legally designates an AI vendor a "national security risk" simply for upholding its own ethical safety guidelines against domestic surveillance, and immediately replaces them with a competitor willing to sign a softer contract, is the government securing the homeland, or is it demonstrating that in the AI arms race, corporate ethics are a disqualifying liability? It's a question worth sitting with.
HostThis has been a whirlwind tour of the current state of AI governance, or perhaps, misgovernance. But what are the absolute key takeaways listeners should have from this?
ExpertI think the first is the **Preemption Paradox**. The DOJ's AI Litigation Task Force represents a complete abandonment of traditional conservative federalism. The federal government is weaponizing its power against the states, not to protect individual liberties, but to shield tech monopolies from compliance costs.
HostAbsolutely. And then there's the **Copyright Civil War**. The White House, via its framework, is essentially saying that scraping the internet for AI training is fair use. Meanwhile, Senator Blackburn's bill is saying it's theft. That divergence creates an unnavigable legal minefield for any developer trying to innovate responsibly.
ExpertAnd let's not forget the **Extortion by Infrastructure**. The administration's threat to withhold $21 billion in BEAD broadband funding has already successfully forced states like Louisiana to abandon AI safety legislation. This isn't just policy; it's economic coercion.
HostFinally, the **Ethics Penalty**. The Pentagon's treatment of Anthropic, as we saw in The Conflict Docket, proves that the government, acting as a customer, will actively punish AI companies that attempt to enforce their own safety guardrails. They'll favor vendors like OpenAI who offer more pliable terms. It sends a chilling message to anyone trying to build AI ethically.
ExpertSo, given all this, two questions really stand out to me: Is the current executive-led "light touch" approach genuinely about fostering innovation, or is it primarily designed to protect the business models and competitive advantage of a select few incumbent tech giants?
HostAnd perhaps more importantly, when the federal government uses both legal and financial strong-arm tactics to dictate AI policy, effectively overriding state-level democratic processes, what does that mean for the future of democratic oversight in a rapidly evolving technological landscape?