Joanna Woronkowicz and Doug Noonan (I/O Research, Inc.) present ARTIFACT, a new open evidence base on AI and the creative industries now available in early public beta at https://artifact.io-research.org/.
This project was funded by one of Creative PEC’s Small Research Network Grants. These awards bring together Creative PEC Research Fellows with the aim of promoting research and policy understanding in key areas and supporting initiatives that deliver tangible benefits within academic, policy and practitioner communities. Find out more about the projects here.
A new open-access knowledge base for research on AI in the cultural and creative industries is now live in public beta, developed with support from a Creative PEC Small Research Network Grant. Here is what it does, why we think the field needs it, and how to get involved.
Information about AI in the creative industries is abundant; evidence is harder to find. Rigorous studies exist, but they are scattered across disciplines, publication types and countries, so practitioners, policymakers and researchers often cannot tell what has been done, where it is, or how the findings fit together. ARTIFACT was created to close that gap.
ARTIFACT stands for AI Research and Toolkit Infrastructure for the Arts and Creative Trades. It is a searchable, open-access wiki that brings together research, people, institutions, projects, datasets and events relating to AI and the creative industries. It is live now as a public beta, and it is the first piece of a longer plan to give the field something it has never had: a shared evidence base that grows over time.
Why the evidence lags the technology
AI is changing how cultural value is produced, who captures it, and how it reaches audiences. Research on these changes exists, but it is scattered. Economists study one piece, legal scholars another, media researchers a third, computer scientists a fourth, and creative practitioners hold knowledge that rarely reaches any of those literatures. Each group uses different words for the same things.
The policy conversation does not wait for any of this to be sorted out. It is increasingly shaped by the actors with the best access to data and the strongest commercial reasons to tell a particular story. Independent, public-interest research struggles to build up fast enough to offer an alternative. The result is that decisions about copyright, licensing, competition and cultural funding risk being made on the strength of promotional claims about adoption and inevitability rather than transparent evidence.
Our starting point is that AI can enhance creative processes, widen access, support experimentation and open new business models. But those benefits will not be fairly understood or governed without an independent research base that can track what is happening to creative labour, rights, cultural diversity, market power and audiences. That is an infrastructure problem.
The current beta
In its current form, ARTIFACT is organised around six kinds of research content: publications, people, institutions, projects, datasets and events. If you want to know who has studied AI and music royalties, whether any dataset exists on AI adoption in newsrooms, or which projects are looking at synthetic media in film production, it is designed to be the first place you look.
It is built for a mixed audience. Researchers who need to find existing work, and each other. Policymakers and advisers who need evidence before a consultation closes. Funders trying to see where the gaps are. And people in cultural organisations, unions, guilds and creative businesses who want to connect what they see on the ground to broader patterns.
We are calling it a beta deliberately. The structure is in place, the contents are partial, and several of the features described below are still being built. What exists now is a working foundation.
What we are building towards
Inclusion criteria, with editorial oversight. Trust in an open resource depends on knowing how entries are selected. ARTIFACT’s criteria are being designed with a human editorial board. The proposed criteria admit peer-reviewed research, policy and legal material from recognised bodies, datasets and projects with documented provenance, and reports from outside academic publishing that experts recommend.
A common vocabulary. A lawyer, an economist, a musician and a machine-learning engineer may describe the same licensing arrangement in incompatible terms, and countries classify the creative industries differently. A shared tagging system will link each entry to standard identifiers for publications, people and institutions and tag it by subsector, AI technology, method, geography, or legal issue, so that studies from different disciplines and countries can be read together.
AI-assisted updating. The literature now grows faster than any team can review manually. In a later phase, AI may be used to ingest new research and policy sources as they appear, suggest where each item belongs, and flag emerging clusters, disagreements and gaps.
Human responsibility for interpretation. Expert judgement will remain responsible for interpretation. Every AI-generated classification or summary will be traceable to its sources, editable, and reviewed before publication. AI is used for scale and speed, not to determine what the evidence means.
Why this matters now
Governments are deciding how copyright applies to training data, whether licensing should be mandated or encouraged, what transparency AI developers owe rights holders, and how competition rules should treat platforms that both host and generate content. These decisions are being taken while AI companies and rights holders are still negotiating, and most of the relevant data is held by those parties.
Three practical consequences follow. First, because every entry must meet published criteria and is checked against editorial criteria, policymakers and practitioners will have a reliable place to test claims about what AI is doing to creative work, rather than another source of opinion. Second, because entries will be tagged consistently, a study from one discipline or country can be read alongside one from another, and evidence about music, for example, will not be mistaken for evidence about publishing. Third, because new research is added as it appears, the evidence available when a decision is being made will be current, and where studies disagree or do not yet exist, that will be plainly visible.
Next steps and how to contribute
The beta is released at this stage to invite contributions. If you are a researcher, the quickest contribution is to check whether your own work, data and projects are in the wiki, and to propose the ones that are not. If you work in a sector body, union, collecting society or cultural organisation, tell us what evidence you are missing and what data you might share under the right conditions. If you advise on policy, tell us which questions are landing on your desk.
The aim is not to settle in advance whether AI benefits or harms the creative industries, but to build infrastructure through which that question can be examined rigorously, repeatedly and in public. And the whole thing only works if it stays open, curated and independent enough that researchers, policymakers and creative communities can use it to shape AI rather than merely react to it.
ARTIFACT is available in early public beta at https://artifact.io-research.org/. It is incomplete, and corrections and suggestions are welcome by email to artifact@io-research.org.
Illustration by Alexander Naughton
