Sundance Institute — AI Literacy Initiative

A three-year, product-agnostic AI education programme launched January 2026 with $2m from Google.org's AI Opportunity Fund: free learning content and events on Sundance Collab, a year-long AI Creators Fellowship, a resource hub, and an AI Literacy Alliance with The Gotham, Film Independent and the Creators Coalition on AI. Film-sector oriented and free to participants. It is literacy, not policy — no research or tracking component — and the funder is a party to the argument, which the programme's product-agnostic framing acknowledges without resolving. It is nonetheless the first substantial US arts-sector answer to the question of what practitioners should actually be taught.

Washington University School of Law

Digital replica and right of publicity

The closest thing to a tracker of state laws extending publicity rights to AI-generated digital replicas of voice, likeness and persona: a structured table with effective dates and links to the statutes, last revised June 2026. Nine states listed, which understates the field — Tennessee's ELVIS Act and Washington's 2026 expansion are among the omissions — and it is a curated law-school resource page rather than a maintained tracker. It is still the only free, neutral, structured thing in this lane. For the performer's-eye view of what is moving, the National Association of Voice Actors keeps a running list of the bills it supports, which is advocacy but is organised and current. NAVA: https://navavoices.org/advocacy/nava-policy-support/

C2PA

Licensing and certification infrastructure — a status report

Between 2023 and 2025 a small industry formed around certifying and licensing human-made or lawfully-trained work. Most of it is now gone. Spawning and its Have I Been Trained opt-out registry serve maintenance pages, and its Source.Plus dataset service now redirects to a domain-sale listing. Human Native AI, the training-data marketplace, was acquired by Cloudflare in January 2026 and folded into its crawler-control products — the independent marketplace no longer exists. Fairly Trained, Ed Newton-Rex's certification for models trained on licensed data, is still accepting applications and lists twenty certified entities, but has published nothing since August 2024. Created by Humans, the Authors Guild–backed rights-licensing platform, is live and appears to be the functioning one, though it publishes no data about its catalogue. The pattern is worth stating plainly, because the individual sites do not state it: voluntary certification and opt-out registries have not sustained themselves. Provenance work has survived where it is embedded in infrastructure companies already sell — which is why C2PA persists and Have I Been Trained does not. Anything built on the assumption that a registry will be there in three years should account for this. https://www.fairlytrained.org/ · https://www.createdbyhumans.ai/

Université de Montréal

Participatory AI Atlas

Rashid Mushkani's project at Mila and the Université de Montréal: 131 catalogued participatory AI initiatives across 36 countries, coded by participation tier, method, and where in the AI lifecycle the participation happened. Published with the FAccT 2026 paper "Voices in the Loop"; free, CC BY, versioned. The headline finding is that participation clusters at problem formulation, governance and evaluation, and virtually disappears at training and development — people are consulted about what a system should be for, rarely about how it gets built. Read it for the method and for the structural limit it shares with everything else in this literature: it catalogues initiatives that were participatory, so a decision made with nobody in the room never enters the dataset. The sharpest complement is Papadopoulos et al., "Participation asymmetry in EU AI governance" (Transforming Government, May 2026, open access), which codes who actually submitted to the three EU AI Act consultation phases and finds EU citizen representation falling to 3.96% at the point the legislation was decided.

Carnegie Mellon University

The Arts and AI Observatory — Carnegie Mellon University

The nearest American peer, and a National Endowment for the Arts Research Lab: Carnegie Mellon with a2ru, Creative Generation and Penn State's Art & Design Research Incubator, describing itself as a national research effort tracking arts, AI and the future of creative practice. Two strands — a two-year cross-sector Fieldscan, and a Benchmarking Initiative mapping how higher education in the creative fields is responding. Listed here on the strength of the design rather than the output: at checking, nothing had been published and no cadence was stated. Its centre of gravity is practice and pedagogy — what artists and art schools are actually doing — which is the question this project's map does not answer. Worth following for that reason.

Creative PEC

Creative PEC — Creative Industries Policy and Evidence Centre

The UK's standing evidence infrastructure for the creative industries, with AI, intellectual property and regulation as one of its ten research themes and 150+ reports since 2018. Two things to watch rather than read now. State of the Nations is the recurring economic assessment. The Creative Business Panel is a five-wave longitudinal survey of UK creative businesses running 2025–2028, backed by DBT and DCMS, explicitly including repeat measurement of emerging-technology adoption — which will make it the first real time series on AI use in creative firms. Wave 1 was due in 2026 and had not appeared at checking. This is what a national evidence base for the creative industries looks like. The United States does not have one.

Water & Music

Water & Music AI Tracker

Cherie Hu's music-industry research collective, and the closest thing anywhere to a sector doing this for itself. Two relevant standing trackers: a Music AI Market Tracker covering 130+ tools, models and datasets alongside partnerships, investments, litigation and policy; and a Music AI Ethics Tracker coding what music companies have actually said about artist consent and data transparency. Behind them sits a proprietary database of 1,000+ music-tech companies including 300+ AI tools. The paid membership model ended and the six-year archive — 300+ articles — is now free.

CREATe

CREATe — AI Licence Tracker

The best dataset in its lane and not close. CREATe, the AHRC-designated copyright research centre at Glasgow, tracks commercial licensing agreements between creative content providers and AI developers: 83 agreements at launch, 274 as of 31 January 2026, coverage running from October 2022, with an interactive flow diagram added in June 2026. Published as an open, downloadable sheet with an invitation to report missing deals — which is the correct way to run a dataset in a field where most deals are announced by press release and some are never announced at all.

CNC

CNC — Observatoire de l’intelligence artificielle

The most substantial thing in this section and the least known outside France. The Centre national du cinéma, a statutory agency under the Ministry of Culture, has run a permanent AI measurement apparatus since March 2024: an annual barometer of AI adoption among film, audiovisual and video game professionals, now in its third edition; a quarterly employment note produced with Afdas and Audiens tracking AI's labour impact on named occupations — sound mixers, storyboard artists, dubbing actors; and a recurring survey of public attitudes. Free PDFs, in French. No other country has this. It is worth knowing about less for the French numbers than for the demonstration: a culture agency can measure this transition quarterly, as a standing public function, and the only reason nobody else does is that nobody else has the institution.

Tow Center

Tow Center — AI Deals and Lawsuits

Launched December 2025 by Columbia's Tow Center for Digital Journalism: a live tracker of the lawsuits, licensing agreements and grants running between news publishers and AI companies. Journalism only, which is a limit and also the point — news is the one creative sector where the licensing market has actually formed, and this is the clearest view of it. Free, no cadence stated. Do not confuse it with the Tow Center's older Source Tracker, which is unrelated.

Mishcon de Reya

Mishcon de Reya — Generative AI IP Cases and Policy Tracker

The only free tracker that carries both litigation and policy in one filterable structure, across the US, UK, EU, France, Germany, Canada and India, with a dedicated category for the newspaper cases. That combination is rare enough to be worth the entry. What it does not have is a stated cadence, a last-updated date or a case count, and it is a law firm's tracker, so the organising logic is legal instrument rather than what is at stake. Useful for a fast cross-jurisdictional read; not something to cite as current without checking the underlying item.

IRIS Merlin

European Audiovisual Observatory — IRIS Merlin

A free legal database of media and copyright developments across 41 European countries, running continuously since 1995 and fed by the monthly IRIS newsletter. It is the fastest way to find out what actually happened in a European jurisdiction without reading a law firm's client alert about it. AI is covered as subject matter — an IRIS report on copyright and AI training was published in July 2026, with a second part on copyrightability of AI outputs due later in the year — but there is no AI category and no AI tracker, so this is a general legal-monitoring instrument you use with a search box. The Observatory does run standing trackers; none of them is about AI.

UNESCO

UNESCO — Policy Monitoring Platform and Re|Shaping Policies for Creativity

The only global inventory of cultural policy that exists: 8,000+ policy measures submitted by states party to the 2005 Convention, filterable by country, cultural domain and theme. The February 2026 Re|Shaping Policies for Creativity report is the flagship analysis built on it, covering 120+ countries with the digital environment and AI foregrounded. Two limits worth knowing before you rely on it. The measures are self-reported by governments on a four-year cycle, so this records what states say they have done, not what has happened. And there is a digital-environment filter but no AI category, so AI measures are not separately retrievable — you are searching, not browsing. Read the report for the international frame and the platform for what a specific ministry has committed to. Report: https://www.unesco.org/en/reshaping-creativity-reports

artscouncil.org.uk — Responsible AI Practical Toolkit

Arts Council England — Responsible AI Practical Toolkit

Seven downloadable guides from a twelve-month R&D programme: starting the conversation internally, stakeholder mapping, writing an AI policy, a risk register, a responsible-AI checklist, and pilot templates. Its framing — that this is more a question about people than about technology — is the right one for organisations that are being sold tools before they have decided anything. ACE published AI Technologies and Emerging Forms of Creative Practice in February 2026, analysing roughly £4m of its own investment in AI-related creative projects since 2019 — which makes it, unexpectedly, one of the few pieces of evidence anywhere about what artists funded to work with these tools actually made.

culturehive.co.uk — AI Policy Workbook

Arts Marketing Association — AI Policy Workbook

The freshest institution-facing resource verified, at version 3 as of June 2026 and still under active development. Covers sixteen policy areas including agentic AI, and is structured as a workbook an organisation can actually complete rather than a document to read. UK-oriented but almost entirely portable.

contentauthenticity.org

Content Authenticity Initiative and C2PA

Provenance is the one remedy technology companies broadly endorse, which is both why it is furthest along and why creator organisations treat it as necessary but insufficient. The Content Authenticity Initiative is the readable entry point — explainers, open-source tooling, and a public verification tool where anyone can inspect a file's Content Credentials. Understand the limit before recommending it: provenance metadata is strippable and adoption is uneven, so a missing credential proves nothing at all. Adopting provenance implies no position on training rights.

authorsalliance.org — Generative AI

Authors Alliance — Generative AI

The sharpest available counter-position, and it comes from an author organisation rather than a tech lobby — which is why it belongs here rather than being dismissed. Authors Alliance represents scholars, nonfiction writers and archivists whose interest is in reach and reuse. It rejects the "lawful access" theory that fair use fails when source material was obtained improperly, arguing that fair use concerns the use rather than the conduct, and warns that a lawful-access rule would chill research, journalism and library archiving while advantaging only well-capitalised firms. The blog is where the current thinking is; the resource page itself lags.

humanartistrycampaign.com

Human Artistry Campaign

A coalition of 300+ organisations — launched by the RIAA and the Recording Academy, and now including SAG-AFTRA, AFM, the PROs, and music publishers — organised around seven principles: licensing for training, consent for voice and likeness, no IP exemptions for AI developers, human authorship as the basis of copyright, and mandatory provenance. Its legislative flagship is the NO FAKES Act. Read this as the industry's consolidated position statement, which is what it is. There is no evidence base here and no analysis, and rightsholder-corporate interests sit alongside individual creator interests without the tension being acknowledged.

sagaftra.org — Artificial Intelligence

SAG-AFTRA — Artificial Intelligence

The most operationally detailed union resource anywhere in the creative industries: AI contract bulletins broken out by media type, the Ethovox voice agreement, the Dynamic AI Audio Waiver, and a running bargaining-and-policy timeline. The union's line is consent, compensation and control, with voice and likeness rights strictly opt-in rather than opt-out — a distinction that carries most of the practical weight in any digital-replica regime. Updated continuously.

authorsguild.org — Artificial Intelligence

Authors Guild — Artificial Intelligence

The most practically useful creator-side resource in the field, and the least rhetorical. Model AI contract clauses, best-practice guidance, an explainer on emerging licensing markets, guidance on the Anthropic settlement, and the Human Authored certification and its public database. The Guild's position is pro-licensing rather than anti-AI: consent, credit, compensation, and a functioning market. Worth noting that the Human Authored mark permits AI use for grammar and research, which has drawn substantive criticism that the certification promises more than it verifies.

cisac.org/AI

CISAC — Artificial Intelligence

The international confederation of authors' societies, representing collecting societies in over 100 countries. Its economic study on generative AI's impact on music and audiovisual creators is one of the few attempts to quantify displacement in creative markets rather than infer it. Equally valuable and less noticed: a public archive of CISAC's own policy submissions across jurisdictions, which is an efficient way to read the rights-holder position stated formally rather than in press releases. This is a rights-holder body arguing its members' interests; the data is useful, the conclusions are advocacy.

oecd.ai

OECD.AI Policy Observatory

A living database of 2,494 initiatives across 80+ jurisdictions, searchable by country and instrument. The best single place to see what a given government has actually committed to, and the only free resource that makes cross-national comparison tractable. Creative-sector and copyright policy is thinly represented — this operates at the level of national AI strategy — but for establishing what frame a country is working in, it is the starting point.

hai.stanford.edu — AI Index 2026

Stanford HAI — AI Index Report

The annual statistical baseline for the AI field: capability benchmarks, investment, labour-market effects, public opinion, and a policy and governance chapter covering global rulemaking. Published each spring; the 2026 edition is current. Underlying data is released openly. Be clear about its limits — there is no copyright or creative-industries chapter, and the policy framing is macro (national investment, AI sovereignty). Use it for the denominator, not the argument.

mckoolsmith.com — AI Litigation Tracker

McKool Smith — AI Litigation Tracker

Scoped specifically to generative-AI copyright litigation affecting media and entertainment, which makes it the most relevant per-case status report for this project's purposes — UMG v. Suno, Sony v. Uncharted Labs, Concord v. Anthropic, the Disney/Universal/Warner actions, Andersen v. Stability, the SDNY MDL. Free, no registration. Cadence has slowed from weekly in 2025 to roughly monthly in 2026, so it lags Lee's site on breaking developments while being better organised for review.

chatgptiseatingtheworld.com

Chat GPT Is Eating the World

Edward Lee's site is the single best free resource in this field, and it is not close. Near-daily coverage of AI copyright and tort litigation plus a searchable case tracker covering a searchable case tracker formally launched in April 2026 at the hundred-case milestone and carrying 128 US copyright suits by mid-July 2026, roughly 143 worldwide. Since then Lee has added parallel trackers for AI authorship, shadow-library litigation strategy, web scraping and AI agents, and a master calendar of hearing dates., with claim charts and jurisdiction maps. Filings, orders and rulings usually appear within a day. Two caveats: the tracker is a JavaScript application, so link to the parent site for anything you need to persist, and this is a one-person operation with the fragility that implies.

transparencycoalition.ai/news

Transparency Coalition — AI Legislative Update

A weekly Friday roundup of state legislative activity during session, tracking training-data transparency, synthetic-content disclosure, and deepfake bills across California and a dozen other states. This is the fastest free source for state-level movement on the questions this project cares about. It is also an advocacy organisation campaigning for transparency legislation — the bill status is reliable, the framing is not neutral, and it should be read as a participant's account.

ncsl.org — AI Legislation Database

NCSL Artificial Intelligence Legislation Database

Free, neutral, monthly, and comprehensive across all fifty states — the baseline for anyone asking what US states have actually enacted. The limitation is taxonomic: it organises by broad AI policy category, so likeness, digital-replica and training-transparency bills are scattered rather than grouped. Expect to filter. The URL path is misleading; this is the live product, and the older "2025 legislation" page is frozen.

wipo.int — Frontier Technologies

WIPO — Frontier Technologies and Intellectual Property

The only forum where all jurisdictions are in the same room. WIPO cannot make rules, and nothing it produces binds anyone, but its convenings surface how differently the same question is being answered across countries. The Twelfth Session of the WIPO Conversation (October 2025) was on IP and synthetic media — deepfakes, unauthorised use of works and likenesses, watermarking and labelling. The IP policy toolkits are genuinely usable by non-lawyers.

gov.uk — Report on Copyright and Artificial Intelligence

UK Report on Copyright and Artificial Intelligence (March 2026)

The most consequential negative result in the field. After the December 2024 consultation, the UK government abandoned its own preferred option — a broad text-and-data-mining exception with a rights-holder opt-out — in the face of creative-sector opposition, found no consensus, and chose no immediate legislative reform. It will instead watch EU transparency implementation and US fair-use litigation. Worth reading closely for what it flags as unfinished: digital replicas and personality rights, protection for computer-generated works, and enforcement.

artificialintelligenceact.eu

EU AI Act Explorer

The searchable full text of the AI Act with an implementation timeline, plain-language article summaries, and a compliance checker — far more navigable than the legislation itself. Maintained by the Future of Life Institute, which is an advocacy organisation, so treat the framing accordingly; the underlying text is the text. Full applicability arrives 2 August 2026, and the 2026 Digital Omnibus amendments shifted several high-risk deadlines.

Copyright.gov/ai

U.S. Copyright Office — Copyright and Artificial Intelligence

The Office's three-part study is the closest thing the United States has to an official position: Part 1 on digital replicas (July 2024), Part 2 on copyrightability (January 2025), Part 3 on generative AI training (May 2025). Read it as the canonical reference, not as a live feed — Part 3 remains a "pre-publication version" more than a year on, and the initiative has not moved since the leadership disruption of 2025. The registration guidance and Review Board decisions on AI-assisted works are the practically useful material here.