Governance

Copyright Infrastructure Task Force (CITF)

Several participants in the emerging ReactAI research and innovation partnership are members of the Copyright Infrastructure Task Force (CITF).

[PLACEHOLDER: This section will be developed with the participating CITF members to describe relevant CITF activities, areas of alignment with the ReactAI Governance Agenda, and appropriate mechanisms through which ReactAI research, implementation experience and demonstrators may contribute to or learn from CITF work. No institutional relationship or endorsement by CITF is implied.]

Federated Copyright & Cultural Data Infrastructures

ReactAI investigates the legal basis and governance framework through which federated copyright and cultural data infrastructures can carry AI-assisted claims while preserving institutional autonomy, legal certainty and human accountability.

A cultural or creative asset may be represented through assertions concerning authorship, ownership, rights, territory, copyright term, commercial availability, permissions and other legally consequential matters. Such assertions may be reconstructed from dispersed documentary evidence and may involve both computational and human processes. They are nevertheless propositions on which another person or organisation may subsequently act and which may be incomplete, contested or wrong.

The governance problem therefore extends beyond the accuracy of an individual assertion. ReactAI investigates the allocation of authority, responsibility and potential liability across the complete claim lifecycle, including the evidence holder, claim generator, validator, issuer and relying party. These roles need not belong to the same organisation, and different organisations may possess different legal mandates and evidentiary authority.

A central research question is consequently whether and under what conditions an institution may assert, validate, issue or rely upon rights-related knowledge concerning an asset that it holds, documents or administers but does not necessarily own. ReactAI also investigates the legal character of validation performed by a party without standing in the underlying right, the warranties or qualifications that an issuing institution can responsibly provide, and the conditions under which another organisation may reasonably rely upon the resulting assertion.

The research considers governance architectures in which no single participant is necessarily the authoritative source for the complete rights status of an asset. Rather than resolving this through centralisation, ReactAI investigates contractual and institutional arrangements through which independently governed participants can contribute different forms of evidence, assertions, validation and authority.

The intended outputs include reusable governance models and legal instruments capable of supporting federated copyright and cultural data infrastructures, including model contribution arrangements, reliance conditions and, where appropriate, analysis of suitable institutional or contractual forms for federation.

Machine-Actionable Rights & Interoperability

ReactAI investigates the legal and operational requirements that machine-actionable rights assertions must satisfy before they can responsibly support discovery, rights management, licensing, access and reuse.

The research starts from legal and operational requirements rather than from a preferred technical standard. It asks what a trustworthy rights assertion must communicate about its source, authority, evidentiary basis, jurisdiction, temporal and territorial scope, validation status and conditions of reliance, and then evaluates whether existing standards and specifications can represent these requirements adequately.

This includes research on machine-actionable policy expression, particularly ODRL and relevant W3C specifications, as well as interoperability requirements emerging from copyright infrastructure work and persistent identifier systems. Existing and emerging identifiers can be evaluated according to the functions they perform within the claim lifecycle rather than treated as interchangeable technical identifiers.

Particular attention is given to the relationship between technical trust and legal or evidentiary authority. Attribution, signatures, timestamps, verifiable credentials, electronic seals and other technical mechanisms may establish important properties of an assertion, but ReactAI investigates what these mechanisms establish legally and what additional institutional authority, evidence or professional judgement may still be required.

Rights assertions are also necessarily contextual. Copyright protection and permissible uses can depend on jurisdiction, territory, time, asset type, contractual arrangements and the particular right or action under consideration. ReactAI therefore investigates representations capable of expressing qualified rights information rather than reducing complex legal states to universal labels such as “in copyright” or “public domain”.

The objective is not to encode European copyright law as a universal computational rule system. It is to determine which legally and operationally relevant information can be represented interoperably and machine-actionably while keeping accountable legal and professional determinations attributable to the competent actors who make them.

Institutional & Operational Validation

Governance architectures cannot be validated through technical reference implementations alone. ReactAI investigates whether proposed legal, governance and interoperability models remain workable when applied by the institutions that would actually contribute, validate, issue, exchange or rely upon rights and cultural knowledge.

Institutional validation examines whether the allocation of authority and responsibility represented by a technical architecture corresponds to the legal mandates, professional responsibilities and operational practices of participating organisations. It also examines whether proposed trust mechanisms, contribution procedures, reliance conditions and machine-actionable rights expressions provide sufficient information for institutions to make accountable decisions.

This research is deliberately operational. Governance assumptions are tested against functioning copyright, cultural heritage and data infrastructures wherever appropriate, rather than only against abstract models. Differences between institutional mandates, national legal environments, professional practices and technical infrastructures are treated as evidence about the limits and requirements of the governance model rather than as deviations that must automatically be harmonised.

ReactAI therefore uses institutional and operational validation to distinguish what can be generalised from what must remain jurisdiction-, sector-, infrastructure- or institution-specific. Where a common European representation or governance pattern is possible, the research seeks to identify it. Where responsible generalisation is not possible, the dependency should remain explicit and configurable.

Relevant public institutions, cultural infrastructures, rights organisations, industry actors and civil-society organisations can participate as reviewers and validation partners according to their actual competence and mandate. Participation in validation does not imply institutional endorsement of ReactAI or of conclusions beyond the scope of that participation.

Implementation & Governance Instruments

ReactAI translates legal and governance research into instruments that organisations can evaluate, adopt, test and implement.

The objective is to move from abstract principles of trustworthy AI, copyright interoperability and federated governance towards explicit arrangements governing how evidence and assertions are contributed, validated, issued, exchanged and relied upon. Where responsibilities can be expressed contractually or procedurally, ReactAI investigates how those responsibilities can be made sufficiently explicit for operational use.

Research outputs may include model contribution terms, model reliance terms, implementation profiles, conformance criteria, trust models and governance guidance. These instruments connect legal analysis with the technical and semantic architectures developed elsewhere in ReactAI while preserving the distinction between technical conformance and legal or institutional authority.

Implementation profiles specify the subset of concepts, representations, provenance requirements and operational rules required for a defined use case or infrastructure. Conformance criteria make it possible to determine whether an implementation satisfies those requirements. Governance guidance describes the institutional responsibilities and decision points that cannot appropriately be reduced to technical conformance alone.

Model contribution and reliance arrangements address the relationships between organisations participating in federated infrastructures. They can specify what a contributor represents about the origin, authority and quality of a contribution; what validation has or has not occurred; what responsibilities remain with the issuer; what a relying party may reasonably infer from an assertion; and how corrections, disputes, withdrawal or new evidence should be handled.

The resulting instruments are intended to make federation implementable without requiring participating organisations to surrender their institutional autonomy or accept undefined responsibilities. They also provide a mechanism through which findings from operational implementations can be translated into reusable governance practice, interoperability specifications and, where appropriate, recommendations for European copyright and cultural data infrastructures.