The previous architecture has been implemented as a PoC in the context of the ONTOCHAIN NGI H2020 project. It is possible to follow the development process, get access to the source code, and test the current version of the application from the project’s dashboard.
9 The dashboard also includes the automated tests used to evaluate CopyrightLY compliance with the requirements specification. The following subsections highlight some of the most relevant parts of the project, including the incentive mechanisms to curate authorship claims and the implementation of licensing NFTs based on semantic technologies and ontologies, which enable copyright reasoning on top of NFTs ownership and tradeability mechanisms.
4.1 Incentives to Curate Authorship Claims
One of the biggest issues detected in the state-of-the-art proposals analyzed in Section
2 is that registering content on-chain does not provide any guarantee of its validity. It might be just the action of someone with access to a digital copy of a creation, but not necessarily the original creator. To deal with these false authorship claims, CopyrightLY leverages two incentive mechanisms. The first one is based on the fact that all claims and evidence are recorded on-chain and can be used in case of litigation. This is possible even in court, like in the TikTok vs. Baidu case [
24], and disincentivizes false claims. However, it just works in the longer term and when some identifying information has been provided to strengthen the claim.
The second incentive mechanism aims at working on the shorter-term, scale thanks to the wisdom of the crowds and incentive stronger authorship claims and complaints, supported by evidence and some identifying information like social media accounts linked to the user identity. The approach is based on a Token Curated Registry [
18], where a token with economic value, the CLY token, is used as an incentive mechanism [
35] to curate the most reliable manifestations for authorship claims and complaints in a decentralized way.
The mechanism requires creators to stake an amount of the CLY token together with their authorship claims, as shown in the upper left part of Figure
3, which depicts CLY tokenomics. Following the “skin in the game” heuristic [
30], creators risk losing their stake when other creators complain about existing claims, as shown in the upper right part of the same figure. Like original claimers, they can also provide supporting evidence to convince other users to support their position with additional stake. In case of conflicting claims, stakes cannot be removed till it is solved and the side with more stake wins the stake of the other side and shares it proportionally to their stake.
Additionally, filing a complaint on an existing authorship claim blocks the ability of its creator to continue monetizing it until the conflict is solved. Creators can start minting NFTs to license their copyright after staking some CLY and a time has passed without complaints, for instance, one week. On the other hand, to avoid spam complaints that look to block monetization opportunities, they become blocking just if the amount of CLY staken on the complaint exceeds that on the challenged manifestation.
To avoid players with big CLY token stakes dominating the game, an appeal mechanism based on an external arbitration court and decentralized justice [
2] is also considered, as shown in the lower right part of Figure
3. In this case, any of the sides can appeal before the deadline to solve a complaint. The appealing side brings the case to the decentralized arbitration system Kleros using its dispute resolution integration plan.
10A court is opened on-chain with links to all evidence accumulated by both sides plus any identifying credentials that the parties might provide. Systems like Kleros are open to jurors that are invited to decide on cases. There is also a token, in the case of Kleros it is called
Pinakion (
PNK), which jurors have to stake and which is used to incentive participation and avoid attacks.
11 In case of a complaint, the final outcome is that the winning side, that with the bigger CLY stake after the one-week deadline or the one winning the appeal, gets all the accumulated staked which is then split among the winners proportionally to their original stake.
Finally, to motivate participation in this staking game, the CLY token has an inflationary behavior [
32] tied to its use. The token price, in Ether, follows a bonding curve [
34] and grows with the token supply. The bonding curve is defined using the Bancor formula [
16] depicted in Equation (
1) and producing a price curve like that shown in Figure
4, which corresponds to an initially very conservative reserve ratio of
\(\frac{1}{2}\) .
Consequently, if the CLY token succeeds and is used more, which requires minting it to increase the token supply, its price also grows. Early adopters can thus benefit from future players entering the game and unstake and burn their CLY for an amount of Ether bigger than what they originally invested. However, to avoid mere speculation, it is important to note that CLY can just be minted to be staked on a manifestation or complaint. It cannot be transferred, just unstaked and burnt in exchange for Ether at the price defined by the current total token supply. Therefore, stakers should take care of doing so on trustful claims that do not put them at risk, based on a reputation system that is built up in a decentralized way and for which the CLY staked is the main signal. In addition to that, each staker has access to all the information accumulated on-chain, including the authorship claim and the evidence accumulated, but also previous claims and complaints,… even the previous performance of the other stakers participating.
4.2 Semantic NFTs and Copyright Reasoning
The main contribution of CopyrightLY is that authorship claims can be monetized using Semantic NFTs, which solve the main issues identified in the state-of-the-art Section
2. First of all, NFTs are linked to manifestations and thus rooted in an authorship claim accompanied by evidence, a commitment in the form of staked CLY and potentially curated as previously detailed.
Additionally, NFTs’ metadata define the scope of the licensed rights, which are modelled using standard semantic technologies, like the
Web Ontology Language (
OWL) [
17], and based on the Copyright Ontology [
13]. They provide the necessary and unambiguous building blocks that also make the terms machine-actionable.
As shown in the architecture, Figure
2, the License NFT smart contract registers them on-chain and stores their metadata on decentralized storage based on IPFS. The licensing terms are serialized as semantic data based on the JSON-LD
12 standard and stored on IPFS resulting in a content hash. This hash is used as the NFT token URI, used to point to the token metadata. The token URI is immutable so it can be verified that the terms have not changed after the token is minted.
The JSON-LD serialization is configured so the ERC-721 standard metadata fields (
name,
description,
external_link,
image, and
animation_url) are available as plain JSON. This way, Semantic NFTs’ metadata is compatible with existing NFT marketplaces. For instance, Listing 1 shows the metadata for an NFT with the name “Reuse license for “Copyright Blockchain””, which corresponds to a piece of video art generated by an Artificial Intelligence, concretely VQGAN+CLIP [
6]. Then, Figure
7 shows that Semantic NFT in an existing NFT marketplace, so ERC-721 interoperability with existing metaverses and marketplaces is retained by Semantic NFTs.
The metadata includes a
description field that describes the licensing terms as plain text for user consumption. The rest of the metadata is tied to different vocabularies and ontologies to enable an unambiguous and rich representation of the terms, which makes them machine-actionable. First of all, the Schema.org vocabulary [
14] is used for the NFT standard fields that have the same meaning in Schema.org, i.e.,
name,
description, and
image. The rest of the metadata is responsible for capturing the license terms that grant certain uses of the copyrighted work to the NFT owner.
For the copyright-related part, terms from the Copyright Ontology
13 are used. They are tied to the
cro prefix in the
@context at the beginning of the JSON-LD serialization, Listing 1 line 4. Overall, the Copyright Ontology defines the following models:
•
Creation Model: the different shapes of copyright creations along their lifecycle, i.e., Manifestation, Performance, Recording…. In the case of Semantic NFTs, authorship claims are Manifestations, i.e., expression of the work into some kind of object (digital or not) that can be used to claim copyright.
—
Example: for the sample Semantic NFT in Listing 1, this corresponds to lines 30–33, where it is defined as the manifestation for which actions are granted by the NFT license.
•
Actions Model: captures the copyright actions moving creations along their lifecycle (
Manifest,
Perform,
Communicate…) together with actions’ dimensions (
who,
what,
when,
where…) and other management actions like
Agree or
Disagree. The approach is similar to Schema.org Actions [
4] and many of the Copyright Ontology terms are aligned or directly reuse Schema.org ones.
—
Example: the actions modeled in the Semantic NFT example are
Agree and
MakeAvailable, lines 11 and 23, respectively. Both are connected with
who performs the action. In the first case, the creator granting the license is represented as a Decentralized Identifier (DID
14) for the Ethereum account used by the creator as shown in line 19. In the second case, anyone owning the license NFT can make the associated content available. The NFT is also identified using a DID, shown in lines 10 and 27. The NFT metadata also details other conditions for the reuse, including a start and end time (lines 24 and 25), an instrument (line 36, which points to the Voxels metaverse Web application), or the authorized territories (lines 39 and 40, Voxels’ neighborhood “The Center” and the “Vibes” island).
•
Rights Model: this part of the Copyright Ontology represents the legal constructs regulating what actions are favored or restricted. Different legal systems can be represented, from generic ones geared towards worldwide harmonization like those proposed by the WIPO [
33] to those in particular legal regimes. The Semantic NFT does not use any term from the Rights Model because what is licensed are specific actions and no rights are transferred.
Based on the semantic data in NFTs, and in the context of the Rights Module introduced as part of the architecture in Section
3.1, it is possible to make the licensing terms machine-actionable. The rights module fetches the semantic data from minted NFTs and monitors on-chain events to track ownership. All this information is combined into a semantic repository with reasoning and semantic query capabilities. Reasoning has been implemented using SPARQL [
11] queries that check if the different dimensions of an intended reuse (who, what, when, where…) do fit within the licensing terms as illustrated in Figure
5.
Matches can be direct, like that the content or data to be reused has a particular identifier (or content hash), but they can also be indirect and require some reasoning. For instance, if the territory where the intended action would take place is part or not of those where it is authorized. The SPARQL queries implementing copyright reasoning are hidden inside the Rights Module, which offers an API providing higher-level operations, like checking if an intended reuse action is allowed by the terms licensed to the user making the request.
Example: An example of a SPARQL query used by the Rights Module is presented in Listing 2. This query is used to check if a given user is authorized to reuse some specific content in a specific metaverse location, in this case, some specific coordinates of the Voxels metaverse. Based on the metadata for the NFT in Listing 1, and if the intended reuser is the owner of the NFT, it will respond affirmatively to the reuse request and include a link to the agreement granting it, i.e., the DID of the license NFT. In this particular case, the Voxels metaverse location where the reuse is intended is coordinates 6510W,100N,
15 which correspond to Voxels’ parcel 4962
16 named “Generative Artworks Gallery”. Though the NFT metadata explicitly authorizes the reuse for Voxel’s The Center
17 neighborhood and the Vibes island,
18 the query implements all the reasoning about how territories are organized in Voxels metaverse. All that information has been captured using Voxels API and modeled as an ontology available online.
19 When the license NFT and ownership information are combined with the ontology about Voxels’ territories and their “part of” relationships, the Rights Module is capable of determining that coordinates “6510W,100N”, i.e., (−65.1 1), are located on parcel 4962 which belongs to the Vibes island, so that is also an authorized location and the user can display the NFT there.