The Direct Answer: What the NO FAKES Act Is
The NO FAKES Act — formally the Nurture Originals, Foster Art, and Keep Entertainment Safe Act — is proposed United States federal legislation designed to give individuals a federal property right in their own voice and likeness, specifically targeting unauthorized digital replicas created with artificial intelligence. First introduced in Congress in 2023 and reintroduced on a bipartisan basis by Senators Maria Cantwell, Marsha Blackburn, Chris Coons, and Amy Klobuchar alongside House colleagues including Representatives Salazar, Dean, and Coons, the bill creates a new cause of action against anyone who produces or distributes an unauthorized AI-generated replica of a person's voice or visual likeness. As of August 2026, the bill advanced through the Senate Judiciary Committee with a unanimous committee vote, a rare show of bipartisan agreement that has kept it alive as one of the most closely watched pieces of AI legislation in Washington.
Also worth reading: What are the current copyright laws and legal risks surrounding AI voice cloning in 2026? · What is a voice cloning consent form template and how should creators use it legally? · What does a complete AI voice cloning compliance checklist look like in 2026?
For voice cloning specifically, the practical effect is this: cloning your own voice remains fully legal and increasingly useful for creators, but cloning someone else's voice without their consent — even for parody-adjacent content, fan projects, or commercial audio work — would expose you to statutory damages, injunctive relief, and potential liability extending to platforms that knowingly host the material. The bill does not ban voice cloning technology. It bans non-consensual use of it.
Why the Bill Exists: The Deepfake Problem in Numbers
The legislative push traces back to a wave of high-profile incidents between 2023 and 2025. An AI-generated song using cloned voices of major artists spread across streaming platforms before being pulled, demonstrating that current copyright law protects sound recordings but not the human voice itself. Scams using cloned voices of family members, executives, and public figures proliferated; the FTC reported a sharp rise in impersonation complaints tied to voice cloning tools that can produce convincing replicas from as little as three seconds of source audio. Voice actors reported losing auditions to synthetic versions of their own performances, sometimes cloned from prior work without notification or payment.
The Human Artistry Campaign, a coalition of music industry organizations, artist groups, and technology companies, has been the most visible advocacy force behind the bill, arguing that a performer's voice is their livelihood and deserves protection analogous to publicity rights. Notably, some voice-cloning companies have endorsed the effort rather than opposed it. Respeecher, a professional voice synthesis company whose technology has been used in film restoration and other licensed applications, publicly endorsed the NO FAKES Act in 2024, positioning consent-based cloning as the industry standard. This matters because it shows the legislation is not anti-AI; it is anti-unauthorized-use.
How the Proposed Law Would Actually Work
Under the bill as drafted, every individual — not just celebrities — would hold a right against the unauthorized production or distribution of a digital replica of their voice or likeness. Key mechanics include:
The right is a property right, meaning it survives death (for a defined term, generally 70 years after death for deceased individuals, with provisions for successors), can be licensed exclusively or non-exclusively, and can be transferred. Licensing agreements must be in writing and must describe the intended uses of the replica with reasonable specificity. There are carve-outs for First Amendment-protected activity: news reporting, commentary, criticism, satire, parody, and documentary uses receive safe-harbor treatment, though the boundaries of those exceptions remain contested among legal commentators.
Remedies include actual damages, profits attributable to the infringement, or statutory damages (with figures discussed in the range of $5,000 per violation, subject to change as the bill moves through markup). Platforms get a notice-and-takedown framework modeled loosely on the DMCA: if a service receives proper notice that hosted content contains an unauthorized digital replica, removing it provides safe harbor from secondary liability. Ignoring valid notices opens the platform to damages claims.
Critics raise legitimate concerns. Free-speech advocates worry the parody and commentary exceptions are narrower than they should be, potentially chilling legitimate creative work. Some tech policy groups argue the property-right structure could create litigation risk for open-source voice tools and hobbyist projects. These objections are real and unresolved, and anyone working in AI audio should track amendments rather than assume the introduced text becomes final law.
NO FAKES Act vs. Existing State Laws
Federal legislation would not exist in a vacuum. Several states already regulate digital replicas, and understanding the patchwork helps clarify what NO FAKES would change.
| Feature | State Laws (e.g., Tennessee ELVIS Act, California AB 1836/2602) | Proposed Federal NO FAKES Act |
|---|---|---|
| Scope | Varies by state; many cover only commercial use or only deceased individuals | Uniform national standard covering all individuals, living and deceased |
| Voice protection | Tennessee's ELVIS Act (signed March 2024) explicitly covers voice; most states cover likeness only | Explicitly covers both voice and likeness as digital replicas |
| Right type | Mostly right-of-publicity tort claims | Federal intellectual-property-style property right, licensable and transferable |
| Platform liability | Limited or absent | Notice-and-takedown safe harbor with secondary liability for ignored notices |
| Posthumous term | Varies widely (some states 10–100 years) | Generally 70 years post-death, assignable to successors |
| Preemption | None; laws coexist unevenly | Would preempt inconsistent state digital-replica laws |
Practical Steps for Creators Using Voice Cloning Legally
If you clone voices in your workflow, the compliant path is straightforward but requires discipline. First, always obtain written consent from any person whose voice you replicate, and make sure the consent document specifies scope: which projects, which languages, what distribution channels, what time period, and whether derivatives are allowed. A vague email saying "sure, go ahead" will not survive scrutiny under a law requiring written licenses describing uses with specificity.
Second, prefer consent-based platforms. Professional services like Respeecher operate on a licensing model where the voice owner is compensated and documented. Consumer-grade tools vary enormously in how they handle consent; some require upload verification, others do not. Choosing tools with built-in consent workflows reduces both legal exposure and reputational risk.
Third, keep records. Save consent agreements, project briefs, and correspondence showing the voice owner approved each use. If a takedown notice ever arrives, documentation is your first line of defense. Fourth, label synthetic audio where feasible. Disclosure is not yet federally mandated for all content, but several platforms now require AI-content labels, and the FCC has moved on AI-generated voices in robocalls following the February 2024 ruling that AI voice cloning in robocalls is illegal under existing TCPA law — a reminder that sector-specific rules already bite.
Fifth, if you hire voice talent, expect contracts to include explicit digital-replica clauses. Under California-style rules spreading elsewhere, generic grants may be unenforceable, so negotiate replica terms directly and price them fairly.
Common Mistakes That Create Legal Exposure
The most common mistake is assuming "fair use" covers voice cloning. Fair use is a copyright doctrine, and a voice is not a copyrighted work — so fair use offers no defense at all for voice replication itself, though it may protect commentary about a person. Creators who believe adding disclaimers like "AI-generated, not affiliated" immunizes them should understand that disclaimers reduce confusion claims but do not cure lack of consent under the proposed statute.
Second mistake: treating parody as a blanket shield. The bill's parody exception exists, but courts will distinguish genuine parody (commenting on the original) from mere imitation wearing a parody label. A comedic sketch that simply replicates a celebrity's voice reading unrelated jokes may fall outside the exception.
Third: assuming private use is safe. Distribution triggers liability, but publishing to a small audience, a Discord server, or a monetized channel still counts as distribution. Fourth: ignoring platform rules while waiting for federal law. YouTube, Spotify, and TikTok already remove unauthorized voice clones under their own policies, often faster than any court could act. Fifth: cloning deceased artists' voices. Posthumous rights under the bill extend decades past death, and estates have proven aggressive litigants even under current state law.
When to Act: Timeline Through Late 2026
As of August 22, 2026, the NO FAKES Act has cleared a key Senate Judiciary Committee vote unanimously but has not passed either chamber. Realistic scenarios: if it passes during the remainder of the 119th Congress, effective dates typically lag enactment by six to twelve months to allow compliance setup, meaning enforceable federal rules could arrive sometime in 2027. If it stalls in an election cycle, expect reintroduction in the next Congress — the bill has already demonstrated durability across sessions, and state-level momentum (Tennessee, California, New York, Illinois) continues regardless of federal progress.
Practical guidance: do not wait for the effective date. Consent documentation practices take time to build into contracts and pipelines, and platforms are enforcing their own policies today. Conversely, do not panic-abandon legitimate voice cloning — self-cloning, licensed cloning, and clearly consensual projects sit comfortably within every version of the bill drafted so far. The rational posture is audit-and-document: inventory where cloned voices appear in your catalog, confirm consent exists for each, and fix gaps now.
Cost Considerations for Compliant Voice Cloning
Compliance has real costs, though modest ones relative to litigation risk. Licensed voice cloning through professional services typically runs from roughly $50–$500 per project for independent creators up to five-figure fees for studio-grade film and game work, depending on usage scope and exclusivity. Custom voice models built from your own recordings cost nothing beyond the tool subscription — consumer plans commonly range $5–$30 per month, with pro tiers $50–$150 per month offering higher fidelity, more languages, and commercial-use licenses. Legal review of a consent agreement template costs a few hundred dollars once and pays for itself across dozens of projects. Statutory damages under the proposed act, by contrast, start in the thousands per violation, plus attorney fees — making the cheapest compliance step (written consent) also the highest-return one.
The Bottom Line
The NO FAKES Act represents the most serious attempt yet to convert voice and likeness protection from a patchwork of state torts into a single federal property right. Its unanimous Senate Judiciary Committee advancement signals genuine momentum, and its endorsement by both artist coalitions and consent-based voice companies suggests the industry is converging on a consent-first norm regardless of the bill's final fate. For creators, the calculus is simple: cloning your own voice or properly licensed voices is safe and increasingly powerful; cloning anyone else's without documented written permission is a bet against both pending law and existing platform enforcement. Build consent into your workflow now, and the legislation — whatever form it finally takes — becomes a competitive moat rather than a threat.