The Harm Is Real — And the Legislation's Ambition Outstrips It
The case for federal action on AI-generated replicas is not frivolous. Nonconsensual sexual deepfakes, electoral impersonation, and voice-cloned fraud have caused documented harm to real people, and the patchwork of state right-of-publicity laws leaves significant jurisdictional gaps. When a constituent's face is plastered into fabricated pornography or a candidate's voice is cloned to spread disinformation, existing defamation and privacy torts prove slow, expensive, and inconsistently available. Congress is right to look for federal solutions.
But the Nurture Originals, Foster Art, and Keep Entertainment Safe Act (NO FAKES Act, S.4591), which passed the Senate Judiciary Committee on June 18, 2026, by unanimous bipartisan vote, attempts to solve a narrow and genuine harm by creating a sweeping new federal property right — one whose blast radius extends well into protected political speech, journalism, parody, and AI development tools that have nothing to do with nonconsensual deepfakes.
What the Bill Actually Creates
The NO FAKES Act, introduced in revised form in May 2026 by Senators Marsha Blackburn (R-Tenn.) and Chris Coons (D-Del.) alongside Representatives Salazar and Dean, would establish a federal property right in any individual's voice, visual likeness, and — critically — general style. This right would be inheritable by heirs, transferable by contract, and enforceable through a DMCA-style notice-and-takedown regime applicable to online platforms. Violations would carry penalties of up to $750,000 per infringing work.
The bill's enforcement architecture mirrors the Digital Millennium Copyright Act's Section 512 framework: a platform receives a takedown notice, removes the content within a defined window, and achieves safe harbor if it complies. The obvious lesson from 25 years of DMCA operation — that this system is routinely weaponised against legitimate speech — appears not to have meaningfully informed the bill's design.
The 'General Style' Problem
No provision of the bill has drawn sharper criticism than the protection extended to a person's general style. This is not a narrowly defined legal term. It does not map cleanly onto copyright's originality doctrine, trademark's likelihood-of-confusion test, or right-of-publicity's traditional focus on commercial exploitation of a recognizable name or face.
For AI development specifically, this is a landmine. Training a text-to-speech model on public speeches, producing a satirical sketch that mimics a politician's rhetorical tics, or building a research tool that analyzes public figures' communication styles could all plausibly be swept within a "general style" right — depending on how the first litigants frame their claims and how federal courts interpret novel statutory language with no interpretive history.
As the Electronic Frontier Foundation warned in June 2026, the bill would effectively give individuals a "heckler's veto" over commentary, criticism, and parody by allowing a notice-and-takedown demand to succeed before any court has reviewed whether the speech was lawful. The burden falls on the platform to investigate and on the speaker to file a counter-notice — a process that has historically chilled smaller publishers and independent creators far more than large, legally resourced platforms.
The Waivable Right That Isn't Really a Right
Perhaps the most underappreciated structural flaw is the bill's treatment of the likeness right as freely transferable. That means the very performers and workers the legislation claims to protect can be contractually required to sign away their rights as a condition of employment. Given the documented power imbalance between entertainment studios and individual actors, musicians, and background performers, this is not a hypothetical edge case — it is standard operating procedure in entertainment contracts.
A federal property right that can be waived before any harm occurs is not a worker protection. It is a mechanism for laundering industry control over performers' identities through the appearance of federal law. The Recording Academy and entertainment industry broadly support the NO FAKES Act, which should be understood not as validation of the bill's worker-protection rationale, but as evidence of whose interests the legislation most reliably serves once the waiver machinery is in place.
The Takedown Regime Will Absorb Legitimate Speech
The Foundation for Individual Rights and Expression (FIRE) has flagged the bill as a real threat to protected expression, and the mechanism is not hard to trace. A satirical deepfake video of a senator delivering an exaggerated version of their actual policy positions would, under the bill's text, constitute an "unauthorized digital replica." The senator — or their estate, decades hence — could issue a takedown notice. The platform faces strong incentives to comply immediately. The video disappears pending resolution, which could take months or years.
This is not an edge case. It describes how political satire, late-night parody, documentary news re-enactments, and academic AI-research demonstrations all function in practice. The bill contains limited carve-outs for parody and satire, but the burden of establishing these exemptions falls on the accused after removal, not on the complaining party before the takedown issues. That inversion is the structural problem, and adding a list of exemptions does not fix it.
What Proportionate Regulation Looks Like
The harm the NO FAKES Act should address is well-defined: non-consensual sexual deepfakes, AI-generated electoral impersonation within defined proximity to elections, and voice-cloned financial fraud. These are specific, documentable categories of harm that justify targeted federal intervention.
A proportionate law would create a narrowly defined civil cause of action for non-consensual sexual deepfakes; establish federal criminal liability for AI-generated electoral impersonation within 90 days of a federal election; require platform labeling of AI-generated content as a transparency measure; and include an explicit, self-executing carve-out that removes parody, satire, journalism, and research from the takedown regime without requiring the speaker to assert and prove the exemption post-removal.
Congress should send the NO FAKES Act back for targeted revision before floor consideration. The unanimous committee vote reflects the political appeal of appearing to act against AI-generated harm — but unanimous committee approval has historically been the signal that substantive objections have not yet been fully heard, not that they have been satisfactorily answered. The genuine harms from AI impersonation deserve a bill that will survive First Amendment scrutiny, not one that trades a real privacy problem for a new speech-suppression mechanism.