Senator Marsha Blackburn’s discussion draft, referred to as the “Trump America AI Act,” attempts something ambitious: solve the federal preemption question and the AI copyright question in a single bill. According to legal analyses of the discussion draft, it would establish a federal duty of care for AI developers while preempting state-level AI development regulations, and it adds a Copyright Act amendment that hasn’t appeared in prior federal AI preemption proposals. The draft is described as seeking to codify the December 2025 AI Preemption Executive Order, per Mintz’s analysis.
The copyright clause is the new element. 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. That’s a significant proposed change. Fair use has been the primary legal theory AI developers rely on to justify training on copyrighted material without licenses. If that theory were foreclosed by statute, not just challenged in court, but affirmatively removed, the training data exposure for AI developers would shift substantially. This provision makes the bill interesting to a completely different set of stakeholders than a preemption-only draft would attract.
The preemption picture is complicated on its own. The Senate voted 99-1 against a proposal that included a state regulation moratorium, per Cooley’s analysis – a result that signals broad bipartisan resistance to sweeping federal override of state AI laws, at least in that form. That vote doesn’t doom the Blackburn draft, but it’s a significant data point about Senate appetite for this category of legislation. The draft is a discussion draft, not yet formally introduced. What that means in practical terms: it’s a conversation starter, not a bill with committee assignments or a markup schedule yet.
The hub’s prior coverage of the two routes to federal AI preemption established the landscape this draft operates in. The Blackburn draft is the named legislative vehicle that prior analysis was anticipating. The copyright clause adds a dimension that wasn’t part of that earlier analysis, and it’s the dimension that most directly affects AI developers’ core operations, not just their compliance posture.
For compliance counsel and in-house legal teams: nothing in this draft requires action today. It’s not enacted, it’s not introduced, and it faces documented congressional resistance on both its major provisions. The monitoring obligation is real, though. Watch for formal introduction (assignment of a bill number), committee referral, and any markup announcements. Watch also for whether a House companion bill emerges. The copyright provision specifically may attract separate legislative attention independent of the preemption debate, rights holders and AI developers are already litigating this question in federal courts, and a statutory resolution, if it came, would affect every AI developer training on web-scraped data.
The deeper issue this draft surfaces: preemption and copyright are politically distinct coalitions. Rights holders who favor the copyright clause may not support preemption. AI companies that favor preemption (wanting to avoid a patchwork of state laws) may not support a fair use carve-out. Combining both in one bill means the opposition is additive. That’s the structural challenge the Blackburn draft faces that a preemption-only bill wouldn’t.