Nothing yet. S. 4591 was advanced out of the Senate Judiciary Committee by unanimous voice vote on 18 June 2026 and has not passed the Senate. If enacted it would create a federal right in voice and visual likeness, add a takedown process built on cryptographic fingerprinting, and make platforms liable for knowingly hosting unauthorised replicas. A character with no real-world counterpart wouldn't be readily identifiable as anyone, so it would sit outside the definition entirely.
Three things make this worth reading now rather than when it passes. The mechanisms are already visible in private contracts. The state-law patchwork it would sit on top of already binds you. And its text has changed once, so what you plan around should be the shape rather than the detail.
Where does the bill actually stand?
S. 4591, introduced in revised form on 20 May 2026, advanced out of the Senate Judiciary Committee by unanimous voice vote on 18 June 2026. It moved to the full Senate. Reporting at the time noted it was unclear when or whether it would reach the Senate floor.
Three Republican senators raised First Amendment concerns at markup without blocking passage, and agreed to keep working with the sponsors. A House companion, H.R. 8915, was referred to House Judiciary.
Our actor-rights guide covers what the bill is. Post 8 covers where its definition draws the line. This post is about what it would change.
What would actually change for you?
Five mechanisms, and they don't affect all creators equally.
| Mechanism | What it would introduce | Who it affects most |
|---|---|---|
| Federal likeness right | A single federal right in voice and visual likeness, replacing reliance on state law alone | Anyone working across state lines, which is most online creators |
| Notice and takedown | A DMCA-modelled process, with removal triggered on notice | Creators who could receive a notice, and those who might need to send one |
| Digital fingerprinting | A cryptographic hash identifying a specific piece of material, so platforms must also remove matching copies | Anyone whose work could be matched against a flagged replica. This is the mechanism nobody discusses |
| Platform liability | Platforms liable for hosting replicas knowing the depicted person didn't consent | Distribution, indirectly. Platforms would tighten their own rules first |
| Non-assignable right | The right couldn't be assigned during life, though it could be licensed | Anyone drafting a buyout. A licence works; a transfer wouldn't |
Source: S. 4591 as advanced from the Senate Judiciary Committee, read September 2026, plus law-firm analysis of committee proceedings. The bill has not passed and its provisions may change.
Where this falls short. Every row describes the text as it stood after committee. A bill that reaches the floor after further negotiation may not carry all five, and the definitions are the part most likely to move.
What is digital fingerprinting, and why does it matter?
The most consequential mechanism in the bill and the least written about.
The bill defines a digital fingerprint as an electronic label or identifier created by a cryptographic hash function, or another process chosen by the online service, that is unique to a specific piece of material such that it's effectively certain the material won't be misidentified as a match for something else.
The takedown process uses it. When a valid notice identifies an unauthorised replica, a platform must remove the identified material — and, for certain providers, also remove all other publicly available instances that match the fingerprint of the replica named in the notice.
So one notice wouldn't take down one upload. It would take down every copy the fingerprint matches, across that service, without a separate notice for each.
For a creator that cuts two ways. If your work is a replica someone objects to, reposts don't help you. If someone is redistributing your work with a replica in it, one notice reaches all of it.
One notice, every matching copy. That's a different removal system from the one we have.
Would it affect characters with no real-world counterpart?
No, and this follows from the definition rather than from an exemption.
The bill covers representations "readily identifiable as the voice or visual likeness of an individual." A character generated from a text description isn't readily identifiable as any individual, because there's no individual. The right has nothing to attach to.
That means the bill wouldn't change most original narrative workflows at all. What it would change is any workflow that puts a real person's likeness into the output — including the case where a "fictional" character was generated from real photographs.
Which is why the practical preparation isn't legal, it's operational. Knowing which of your characters have real-person references in their generation history is the thing that would determine your exposure, and it's knowable today.
What are the exclusions?
Five, and they're broader than most coverage suggests. Post 8 covers them in detail; the summary matters here because they're the reason the bill cleared committee.
News, public affairs and sports where the replica is materially relevant. Documentary, historical and biographical work representing the person as themselves, including some degree of fictionalization. Commentary, criticism, scholarship, satire and parody in the public interest. Incidental or negligible use. And advertising that promotes any of those four.
Two things bound them. The documentary exclusion fails where the use creates a false impression of authenticity or involves unauthorised musical synchronisation. And no exclusion at all applies where a replica depicts sexually explicit conduct — that override is absolute.
Where this falls short. Whether a given piece is commentary or commercial use is exactly the question that gets litigated, and a framing device doesn't convert an advertisement into satire. The exclusions are real and they aren't a workaround.
How long would the right last after death?
This is where the reporting and the text diverge, and the text is more interesting.
Most coverage says the post-mortem right is capped at 70 years. That's true and it's the least useful way to describe it.
The base term is ten years after death. It's renewable in successive five-year periods — but only where the right holder demonstrates active and authorised public use in interstate or foreign commerce during the preceding two-year window. Seventy years is the hard ceiling on all that renewing, not the term.
So the practical position is nearly the reverse of the headline. A post-mortem right would lapse at ten years unless someone is actively and commercially using the likeness and files to extend it. For a well-managed estate that's routine. For most people it isn't, and the right would simply expire.
If you're considering a deceased person's likeness, that's the fact that actually matters — and the one to check rather than assume.
What's the argument against it?
Worth knowing, because it's the reason passage isn't assured.
The Foundation for Individual Rights and Expression opposes the bill on First Amendment grounds. Its argument is that the definition reaches a wide range of expression — fully virtual creations and materially altered real recordings alike — and that most of what's covered is protected speech Congress can't suppress. It acknowledges the exclusions and argues they don't go far enough.
Three Republican senators raised similar concerns at markup and let the bill through anyway, agreeing to keep working with the sponsors.
Whether that's right isn't something a blog post settles. What it tells you is that the definitions are the contested part, and the definitions are what your planning would rest on.
What already applies, regardless?
This is the part most coverage skips, and it's the part that binds you today.
State right-of-publicity laws already exist and already apply. They vary considerably — in scope, in whether they survive death, and in what counts as commercial use. The bill as advanced would preempt future state laws while preserving existing ones, so the patchwork doesn't disappear. It gets a federal layer on top.
Separately, state AI disclosure laws are arriving faster than the federal bill has moved. New York's synthetic performer law requires advertisers to disclose conspicuously when a synthetic performer appears in a visual or audiovisual advertisement — currently the only US law specifically requiring AI disclosure in advertising.
And platform licences already impose more than the bill would. Hexcoded's actor licence, for instance, already requires verified consent, pays per render, permits digital use only, and lets a creator stop new use at any time. None of that waits on legislation.
What should you do now?
Four things, none of which depends on the bill passing.
Know which characters have real-person references
Go through your existing cast and record, per character, whether any real person's photographs, footage or recordings went into generating it. That single fact would determine your exposure under the bill, and under state law today.
Keep the consent you already have retrievable
A signed release you can't find isn't a rights position. Scope, territory, duration and payment history need to be producible, not just filed.
Licence rather than buy out
The right wouldn't be assignable during life. Any paperwork structured as a transfer of likeness rights rather than a licence is building on something the bill wouldn't permit.
Note what created each generated character
Five seconds at the time, unrecoverable afterwards. It's the evidence that a character with no real-world counterpart genuinely has none.
What to watch
- Whether S. 4591 reaches the Senate floor at all
- Whether the definitions survive further negotiation — they've changed once already
- Whether state legislatures keep moving faster than Congress
- Whether platforms tighten their own rules ahead of any liability provision
Current as of the publication date. S. 4591 is pending legislation and had not passed the Senate at the time of writing. Likeness law varies by state. Hexcoded's actor licence is governed by the laws of India, with jurisdiction in Gurugram, Haryana. Nothing in this post is legal advice — check the current bill status and speak to a lawyer before relying on any of it.
- It's a bill. It cleared committee on 18 June 2026 and hasn't passed the Senate. Don't plan around it as law
- If enacted it would create a federal likeness right, a takedown process, and platform liability for knowing hosts
- The takedown runs on cryptographic fingerprinting. One notice would reach every matching copy, not one upload
- Characters with no real-world counterpart aren't readily identifiable as anyone, so the definition wouldn't reach them
- Five exclusions, broader than reported — including incidental use and advertising a permitted use
- Post-mortem rights would last ten years, renewable only on proof of active commercial use. Not "70 years"
- Structure likeness paperwork as a licence, not a transfer. The right wouldn't be assignable during life
- State law already binds you, and state disclosure laws are moving faster than Congress
No. S. 4591 was advanced out of the Senate Judiciary Committee by unanimous voice vote on 18 June 2026 and moved to the full Senate. It hasn't passed. Reporting at the time noted it was unclear when or whether it would reach the Senate floor.
Create a federal right in a person's voice and visual likeness, make distributors liable for unauthorised replicas, make platforms liable for hosting them knowing the depicted person didn't consent, and establish a notice-and-takedown process built on cryptographic fingerprinting.
A cryptographic hash unique to a specific piece of material, such that it's effectively certain the material won't be misidentified as a match for something else. The takedown process uses it — a valid notice would require removal of the identified material and, for certain providers, all other publicly available copies matching the fingerprint.
Not on the definition as drafted. The bill covers representations readily identifiable as the voice or visual likeness of an individual, and a character generated from a text description isn't readily identifiable as anyone. What it would reach is any workflow putting a real person's likeness into the output.
Ten years after death by default, renewable in five-year periods only where the right holder demonstrates active authorised commercial use in the preceding two years, with a hard ceiling of 70 years. So most rights would lapse at ten rather than running to 70 — which is how it's usually reported.
Four things that don't depend on enactment: record which characters have real-person references, keep existing consent retrievable rather than merely filed, structure likeness paperwork as a licence rather than a transfer, and note what created each generated character.
Consent on record, whatever passes
AI actors with no real-world counterpart, or human creators with verified consent and payment on every render. Hexcoded's licence already requires more than the bill would — which is the point of building on a documented position rather than a pending one.
Read the actor licenceMore on characters, consent and reusable elements in Elements.