Verifiable provenance, rights, and AI transparency for the open web

Digital signatures for content – the core innovation

Liccium lets creators and rightsholders digitally sign claims about their work and bind them to the content’s own fingerprint (ISO 24138). The signature ties a declaration to a verifiable identity, so attribution, provenance, and integrity can be checked independently. This is the foundation the other solutions build on – the trust component the web has been missing.

Signatures are backed by X.509 certificates from EU trust services, or by verifiable credentials issued by trusted providers. Whichever route is used, the declaration resolves to a source that can be verified without relying on any single platform.

Creator Credentials – identity you control

Creator Credentials brings decentralised digital identity to the cultural and creative industries. Through it, creators and rightsholders can request, hold, and manage their own verifiable credentials, keeping a decentralised, user-controlled identity that stays pseudonymous where preferred.

The issuing side sits with the organisations creators already belong to. Media organisations, membership associations, collective management organisations, and trust services can issue credentials to their members, acting as trust anchors for attribution and accountability across digital publishing.

FAIA – AI attribution

As AI-generated media becomes hard to distinguish from human-made work, the web needs a reliable way to disclose how content was created. FAIA (FAIR AI Attribution) provides the vocabulary and registry for it: a signed, machine-readable flag – Human-Created, AI-Assisted, or AI-Generated – bound to the content fingerprint.

Because the flag is bound to the content fingerprint rather than to a file’s embedded metadata, anyone can regenerate the fingerprint from the file and resolve the flag – reliably, and even after the content has been shared or modified. This supports transparency and accountability across the content supply chain, and helps meet disclosure requirements such as Article 50 of the EU AI Act.

TDM·AI – AI opt-out

Rightsholders can reserve their content from AI training. TDM·AI binds a machine-readable opt-out preference to the asset’s content fingerprint, so the reservation survives distribution, manipulation, and metadata removal.

That binding to the asset is the crucial difference. Current approaches – the W3C TDM Reservation Protocol (TDMRep) and the emerging IETF AI Preferences (AIPREF) – express preferences at the level of domains and servers, unsigned and unattributable.

Binding the preference to the fingerprint and signing it is the most robust method: the reservation is provable, verifiable, and cannot be silently detached from the work. It builds on Article 4(3) of the EU CDSM Copyright Directive (2019/790).

Federated registries – the infrastructure

Declarations are published to open, federated registries – a network of independently operated directories rather than one central database. The registries hold the declarations, not the media itself, and index each entry by its content fingerprint. Anyone who generates the ISCC from a file can query a registry and retrieve the declaration bound to it – regardless of where the content is hosted or how it has been shared. Rights, provenance, and metadata live in the registry rather than embedded in the file, so they cannot be stripped along with a file’s metadata.

Registries are run independently – by creator groups, collecting societies, standards bodies, or compliance platforms – and synchronise over a peer-to-peer network using a shared schema, so they work as one globally searchable system that scales to billions of assets. The public index carries only what is needed to confirm a declaration’s rights status; access to the full declaration data can be governed by policy. No single authority controls the system, which is what makes it usable as shared infrastructure rather than another platform.