A consented likeness.
A resolvable chain of custody.

In September 2024, an IBC Accelerator project proposed and led by HAND showed — end to end, in a working pipeline — how a human likeness can be created with consent, identified at birth, signed into the content it appears in, and verified where audiences meet it. Two years later, the rules have caught up with the workflow. Most people still haven’t seen it.

AI can copy anyone.
Could the industry prove who’s real — and who consented?

The IBC Accelerator challenge Digital Replicas & Talent ID: Provenance, Verification and New Automated Workflows took on a question every studio, broadcaster, and guild now faces: when a performer’s likeness can be generated by machine, how do you tell an authorized likeness from an unauthorized one — automatically, across every platform it travels to? HAND first proposed the challenge in November 2023 and pitched it at the IBC Accelerator Kickstart Day in London in March 2024. The answer it tested was a standardized, resolvable identity framework for real people, their licensed likenesses, and fictional characters, carried inside the media itself.

Seven results.
All of them still standing.

01
A consented likeness of a working performer
Actor, rapper, DJ, and SAG-AFTRA member Evan Shafran was scanned with his consent, and a video of his Licensed Human Likeness was produced.
02
Two HAND IDs, issued at birth
Evan (Type 01) and his likeness (Type 02) were both registered in the HAND registry — resolvable, unique, interoperable, and linked to each other.
03
Content and talent IDs, cryptographically signed
C2PA Content Credentials for the likeness video carried the EIDR content ID and the HAND talent ID as source metadata, using EZDRM technology.
04
An asset & metadata workflow, greenlight to delivery
A mapped workflow tracking a likeness from project greenlight through capture, creation, and distribution — across theatrical, TV, gaming, and advertising use cases.
05
Connectors to the people who pay talent
Workflow connectors mapped to accounting and finance teams — where consent becomes credit and compensation.
06
Verification at the point of distribution
Content Credentials were detected in the delivered media at distribution end-points — provenance that travels with the file.
07
Deepfake detection alongside
Videntifier’s forensic media analysis demonstrated detection of manipulated media — the counterpart to proving what’s authentic.
→
Read every detail
Workflows, use cases, legal context, and participants are documented in the public IBC project book ↗

“To have a digital replica made of myself, which I own, is amazing. I’m a performer and I’m an artist, and we’re putting our likeness out there. Somebody is making it work for us — but it’s still our image at the end of the day.”

The identity graph from 2024
resolves right now.

This is not a mock-up. These three records sit in the HAND registry today. The likeness points to the human it is licensed from, and to the company that produced it. Click any node to open the live record.

Evan Shafran’s identity graph in the HAND registry A Type 02 Licensed Human Likeness record links to the Type 01 human it is licensed from, via the licensedHuman edge, and to the Type 04 legal entity that produced it, The Scan Truck, via the productionEntity edge. licensedHuman productionEntity TYPE 02 · LICENSED HUMAN LIKENESS Evan Shafran LHL #0001 10.23/E46E-7EB9-3AF8-E551-23D4 ↗ TYPE 01 · HUMAN IDENTITY Evan Shafran 10.23/59BB-1EA7-1BAB-413B-37DC ↗ TYPE 04 · LEGAL ENTITY The Scan Truck 10.23/F90B-E2F3-AF95-276A-A2D7 ↗
And on the video itself: a C2PA Content Credential carrying EIDR · what was made HAND · who is in it

The Four C’s —
in the order that protects people.

The project framed every use of a performer’s name, image, likeness, and voice around four principles — first said publicly by Alexandra Shannon of CAA at HAND’s own “Battling Deepfakes” webinar, co-hosted with CDSA, on February 7, 2024. Their order matters: control exercised only after distribution and payment is remediation, not control.

01
Consent
The performer agrees, on the record, before the likeness exists.
02
Control
The performer governs how, where, and for how long it is used.
03
Credit
Every use is attributable to the right person, by a resolvable ID.
04
Compensation
The result of the first three: the right person gets paid.

The rules caught up
with the workflow.

In 2024 this was a proof of concept. In 2026 it is what contracts and statutes — in California, the EU, and beyond — assume exists. (The law calls these “digital replicas.” HAND’s canonical term is Licensed Human Likeness — and HAND holds that every synthetic likeness of a person deserves consent-based protection, whatever its statutory label.)

Labor
SAG-AFTRA’s 2026 contract
Ratified June 2026 with expanded AI protections, running through June 30, 2030 — consent and compensation terms that need verifiable identity to enforce at scale.
Federal
NO FAKES Act
Cleared the Senate Judiciary Committee unanimously in June 2026 — the closest the U.S. has come to a federal right in voice and likeness.
State
California AB 1836 & AB 2602
In force: consent requirements for digital replicas of performers, living and deceased, in the state where the studios are.
European Union
EU AI Act, Article 50
In force since August 2, 2026. AI-generated audio, image, and video must carry machine-readable marking — with a grace period for systems already on the market that ends December 2, 2026.
Provenance
California: don’t strip it
From January 1, 2027, the California AI Transparency Act requires large online platforms to detect and display provenance data, and bars them from knowingly stripping it — exactly the distribution-end check demonstrated at IBC 2024. Utah’s HB 276 follows the same day.
International
Mexico’s copyright reform
In force since May 15, 2026: express consent for AI-generated uses of a performer’s voice and image, and a prior written agreement for any AI cloning.

Built in the open,
with the industry.

HAND proposed this project and assembled the collaborator team in cooperation with IBC. Organizations are named as credited at the time. With thanks to Evan Shafran.