Personæ · Open Standard

The Consent Standard

What must be true before a real person's face and voice are recreated by a machine — and what must remain true afterwards.

Version 1.0 Published 10 September 2026 Status Open — free to adopt Priority USPTO 64/009,279

Recreating someone's likeness is now trivial. Establishing that they agreed to it is not, and almost nobody is doing it. This document sets out the minimum we hold ourselves to. It is published so that others can hold themselves to it, and so that anyone buying this technology has something concrete to ask a vendor for.

It is deliberately short. A standard nobody reads protects nobody.


01 Consent precedes creation

No likeness is generated before a consent record exists. This is enforced in software, not in policy: provisioning refuses to run without one. There is no override, and no "we'll collect it later."

Consent is given by the depicted person, actively — a checkbox that is never pre-ticked, and a typed signature. Consent given by an employer on an employee's behalf is not consent.

The language, verbatim

English · v1.0

I authorize the creation and use of an AI-generated digital likeness of my face and a synthetic version of my voice, from the photograph and voice sample I provide, to generate video and audio messages on my behalf. I confirm I am the person depicted, that I own or control these rights, and I grant permission to produce and deliver these AI-generated messages to prospective and existing customers. I understand these messages are AI-generated. I may revoke this authorization at any time, which stops future generation; revocation does not retroactively affect messages already sent.

Español · v1.0

Autorizo la creación y el uso de una imagen digital generada por inteligencia artificial de mi rostro y una versión sintética de mi voz, a partir de la foto y la muestra de voz que proporciono, para generar mensajes de video y audio en mi nombre. Confirmo que soy la persona representada, que poseo o controlo estos derechos, y otorgo permiso para producir y entregar estos mensajes generados con IA a clientes actuales y potenciales. Entiendo que estos mensajes son generados por IA. Puedo revocar esta autorización en cualquier momento, lo cual detiene la generación futura; la revocación no afecta los mensajes ya enviados.


02 Disclosure travels with the output

Every generated message carries a visible statement that it is AI-generated and that the depicted person consented. Not in a footer nobody reads — on the artifact itself, wherever it is played.

On every output

AI-generated message from [Name] · created with their consent.
Mensaje generado con IA de parte de [Nombre] · creado con su consentimiento.

A recipient who cannot tell they are watching a synthetic person has been deceived, whatever the paperwork says.


03 The record is the asset

Consent that cannot be produced on demand did not happen. Every record is immutable and versioned, so it can be shown years later exactly as it was agreed.

FieldWhy it is captured
signer_nameTyped signature, under the E-SIGN Act
signer_emailIdentifies the depicted person, not the buyer
timestampProves consent preceded creation
ipCorroborates the signing event
consent_versionProves what was agreed, not merely that something was
scopeBounds the permitted use

Versioning is the clause most often skipped and the one that matters most. Without it, an organization can only say a person consented — never to what.


04 Revocation is real, and it is bounded


05 What this is built against

These are the regimes the standard is designed to satisfy. It is not legal advice, and it is not a substitute for counsel in any jurisdiction.

TN
ELVIS Act — voice and likeness protection, including liability for the tools that distribute it.
NY
Synthetic-performer disclosure requirements.
CA
AI Transparency Act, effective 2 August 2026 — manifest disclosure on generated content. Civil Code §3344 for likeness.
IL
BIPA — treats a voiceprint as a biometric identifier, requires consent before collection, and carries a private right of action. Separately, HB 4875 covers unauthorized AI replicas.
PR
Act 139-2011 — written consent is the standard; a single unauthorized commercial use is a violation, regardless of any employment relationship.
BR
LGPD — biometric data as sensitive personal data.

The pattern across all of them is the same. The line between lawful and unlawful runs through consent — obtained first, recorded properly, and disclosed at the point of delivery. Everything in this document follows from that.


06 Adoption

This standard is free to adopt, in whole or in part, with or without attribution. Copy the clauses. Use the field list. Hold your vendors to it.

It is published by Personæ, which builds AI-generated digital identities for organizations, and which therefore has every commercial reason to want this settled rather than argued.

Underlying method: USPTO provisional application 64/009,279Composable Digital Identity Marketplace and Licensing Platform, filed 18 March 2026. Publication of this standard is not a licence to the claimed method; the consent language and record format above are offered freely and separately.