We’ve entered an era where anyone can upload a photo of you or a likeness that resembles you and generate anything: videos, images, ads, fake endorsements. It’s frictionless. It’s commercialized. And it’s happening without your consent.
New regulations from the EU AI Act and emerging U.S. frameworks mandate either: Explicit user consent for the use of biometric data (like faces), or Complete exclusion of such identities from generative models. Penalty: Up to €35 million or 7% of global revenue, with enforcement beginning August 2025.
<aside> 💡
In this new AI reality, we need neutral infrastructure not governed by states or the goodwill of a single company, but by individuals themselves. A programmable, decentralized layer where you set the rule.
The Digital Likeness Registry is а solution built on Rarimo, that lets people:
Rarimo is ahead in making it real. We co-authored ERC-7812 with Vitalik to enable on-chain private ZK registries, and unlocked zkML for face processing on-device. With proof optimizations that make checks cheap and scalable, we’ve built the foundation layer to let users register their likeness, enforce rules, or monetize.
To simplify the construction, let’s say each record in the registry consists of two parts: a key and a value (actually, the registry is merkelized to reduce the cost of generating membership and non-membership proofs).
As a key, we use an extraction from user biometric data. As a value, we store policies defined by the biometry owner and additional required information.
Each user can add a record to the registry. For this, users need to pass their biometric data through the extractor, receive the unique value, define access policies (such as “PROHIBITED”, “FREE USAGE”, “PAYABLE”, etc.), and attach additional data if required (for example, an account address for payable entries).