One Bill, Two Separate Political Fights
Most federal AI legislation proposals focus on a single fault line, either the question of whether federal law should displace state AI regulations, or the question of what AI companies owe to copyright holders. Senator Blackburn’s discussion draft, referred to as the “Trump America AI Act,” attempts to resolve both at once. According to legal analyses of the discussion draft, the bill would establish a federal duty of care for AI developers, preempt state-level AI development regulations, and amend the Copyright Act to explicitly exclude unauthorized computational processing for AI training from fair use protections.
Combining these two provisions isn’t just an editorial choice. It’s a structural decision with significant political consequences. To understand why, you have to map the stakeholders independently, because the preemption fight and the copyright fight involve different industries, different congressional relationships, and different legal theories. They share a bill number, but they don’t share a coalition.
The Preemption Provision: Who Wants It, Who Doesn’t
The preemption provision has a clear primary beneficiary: large AI developers operating nationally. A patchwork of state AI laws, with different definitions of high-risk AI, different documentation requirements, different enforcement mechanisms, creates compliance costs that scale with the number of states. A single federal standard is operationally simpler, even if the standard itself is demanding.
The draft is described by Mintz’s legal analysis as seeking to codify the December 2025 AI Preemption Executive Order, which established the federal government’s preference for a unified national AI regulatory framework. The executive branch position is clear. The legislative branch position is not.
The Senate voted 99-1 against a proposal that included a state regulation moratorium, per Cooley’s analysis. That vote is the most important single data point in this story. A 99-1 result isn’t a close call. It isn’t a narrow coalition disagreement. It signals near-universal Senate resistance to the principle of federally overriding state AI authority, at least as that principle was framed in the prior proposal. Whether the Blackburn draft’s framing is different enough to move that number is the central legislative question.
The opposition to preemption is structurally diverse. State governments – including those that have invested significantly in AI-specific legislation – have a direct institutional interest in preserving their regulatory authority. Consumer and civil rights organizations prefer state law flexibility because state legislatures have moved faster than Congress on specific AI harms. Some corporate interests that might otherwise favor preemption have opposed specific federal frameworks that they view as more burdensome than the state laws they’d replace. The 99-1 vote captures all of these positions in a single number.
The Copyright Provision: A Separate Battle with Its Own Coalition
The copyright clause introduces a completely different set of stakeholders. According to legal analyses of the discussion draft, the bill would amend the Copyright Act to explicitly exclude unauthorized computational processing for AI training from fair use protections. To understand what that means, consider the current state of AI copyright litigation.
Fair use has been the primary legal defense AI developers deploy when training on copyrighted material without licenses. Multiple active federal lawsuits, brought by publishers, musicians, visual artists, and news organizations, challenge whether that defense holds. Courts are currently working through those cases with no statutory resolution in sight. A legislative amendment that foreclosed the fair use argument wouldn’t just affect pending litigation. It would alter the fundamental legal basis for how AI systems are trained on web-scraped and licensed content.
Rights holders, record labels, publishers, major entertainment companies, individual creators, have a strong interest in this provision. If enacted, it would give them a statutory basis for licensing negotiations that they don’t currently have as a matter of law. Their coalition includes some of the most effective lobbying organizations in Washington.
AI developers have the opposite interest. A fair use carve-out for AI training would expose every major AI lab’s training data practices to licensing liability. The cost implications aren’t marginal, they’re structural. The AI developers most affected by the preemption provision (which benefits them) are also the AI developers most harmed by the copyright provision (which doesn’t). That’s the internal tension the bill creates for its own potential supporters.
Why Combining the Provisions Creates Additive Opposition
In coalition mathematics, a bill’s opposition equals the union of all groups opposed to any provision, not the intersection. The Blackburn draft’s preemption provision brings in state governments, civil rights groups, and consumer advocates. Its copyright provision brings in AI developers and tech companies. These groups don’t typically share a legislative agenda. Combining their opposition doesn’t create a unified coalition, it just means both coalitions are active simultaneously.
Proponents of the combined approach would argue that the bill needs both provisions to build a sufficient supporting coalition: AI developers accept the copyright clause in exchange for preemption relief, while rights holders accept preemption in exchange for the copyright clause. That theory of the case is possible. It requires both industries to evaluate a mixed-outcome bill rather than a clear-win one, and it requires Congress to hold that coalition together through a markup and floor vote process that the 99-1 result suggests will face significant headwinds regardless of bill structure.
What Compliance Teams and Developers Need to Monitor
Nothing in this draft requires immediate compliance action. It’s a discussion draft, not enacted law, not yet formally introduced. But the monitoring obligations are real, and they’re different depending on your organization’s exposure:
If you’re primarily concerned about the preemption provision (state AI law compliance complexity): Watch for formal introduction and committee referral. The relevant committees are Senate Commerce and Senate Judiciary. Watch for companion legislation in the House. The state law landscape doesn’t change until federal preemption is enacted, continue tracking state-specific requirements, particularly in Colorado, Texas, and California, where AI legislation has been most active.
Note: Current enacted status of specific state AI laws is a content gap in this production run. The Filter flagged this section for Builder source verification, which could not be completed this cycle. The hub’s prior coverage of Colorado’s June deadline provides partial context. A dedicated state AI law status tracker is a recommended content development item.
If you’re primarily concerned about the copyright provision (training data liability): This is the provision that warrants legal counsel review now, not because it’s imminent but because it’s the first time a named federal legislative vehicle has proposed a fair use carve-out for AI training. Understanding your current training data exposure under existing fair use doctrine – before any legislative change, is defensible diligence regardless of this draft’s fate.
For everyone: Track whether the discussion draft becomes a formally introduced bill (assigned a bill number), whether it receives a committee hearing, and whether the copyright and preemption provisions travel together or get separated into distinct bills. Separation would be the signal that the coalition theory isn’t holding.
The Pattern Behind the Draft
The hub has now covered federal AI preemption from three angles in recent weeks: the executive branch pressure on Congress, the legal routes to preemption, and now a named legislative vehicle. The pattern is clear, federal preemption is the dominant legislative question in US AI policy right now, with the copyright dimension emerging as the complicating factor that prior preemption proposals didn’t have to navigate. The Blackburn draft doesn’t resolve that tension. It surfaces it. How the Senate responds to the combined proposal will reveal more about the eventual shape of federal AI law than the draft itself ever could.