Creator-class collective bargaining and litigation on generative-AI training and likeness

Writers, actors, illustrators, voice artists, and authors use the institutions of their professions — guilds, unions, professional associations, trade groups — to negotiate or litigate protections against the unconsented use of their work and likeness to train and deploy generative AI. The instruments are familiar: industry contracts (the WGA / SAG-AFTRA settlements), copyright class actions against model developers, federal-policy advocacy on training-data disclosure, professional-association codes of conduct.

An actor chooses this because creators have an unusually well-developed pre-existing organising substrate for an industrial demand and a clean legal anchor (copyright, right of publicity) that AI training plausibly violates. The fight gives the broader make-AI-good movement what it otherwise lacks: a constituency of named, sympathetic individuals, an immediate economic case the press can carry, and a body of case law that may set the consent-and-compensation default for everyone else's data, not just creators'.

It trades off generality for traction. Wins flow first to the unionised tier — Hollywood writers and actors, big-press authors — and risk leaving non-unionised creators (most illustrators, most freelance journalists) under worse terms after the settlement than before. And copyright is a narrow weapon for what is, structurally, a labour and consent question.

Verdict — most concrete extractive wins in the corpus, bounded by the union perimeter

Creator-class collective bargaining is the corpus's most concrete extractive strategy on AI training-data and likeness — the only strategy that has, on the public record, moved AI from an unbargained-for technology drift in the production pipelines of major frontier-model deployers and adjacent commercial-distribution partners into a set of named, enforceable contract clauses and surviving federal-court claims. The strategy's adopter set is structurally narrow but qualitatively powerful: the Writers Guild of America and SAG-AFTRA on the Hollywood production side, the Authors Guild and the NewsGuild-CWA on the publishing-and-journalism side, the Concept Art Association on the visual-artist side, the WGA / SAG-AFTRA 2023 strikes, Authors Guild v. OpenAI, Andersen v. Stability AI, the Concept Art Association federal-policy advocacy track, and the cross-sector Human Artistry Campaign. The four effects: above are typical of what the strategy returns to the movement — a ratified industry contract whose Article 5 binds generative AI by category, a five-category digital-replica regime that closed the AMPTP's most-aggressive likeness proposal, the first US federal copyright-survival ruling on artists' training-data claims, and the field-defining writer-side AI-copyright decision in a twelve-case consolidated proceeding against the largest US AI developer and its commercial-distribution partner.

The strategy is good for the protected tier, structurally exclusionary for the unprotected creative workforce.

Strong on its primary claim — that a sectoral organising vehicle with existing institutional capacity can extract enforceable AI protections from frontier-AI developers and their commercial-distribution partners faster than the legislative process can. The 2023 Hollywood strikes converted generative AI from a topic the AMPTP had no contractual obligation to bargain over into a contract regime with named consent-and-compensation rules in 148 and 118 days respectively, against a counter-party whose ten-member ownership base controls the entire US studio-and-streamer production system. The Authors Guild moved from the 18 July 2023 open letter to a 20 September 2023 SDNY filing to a consolidated twelve-case master action against OpenAI and Microsoft in roughly two years — faster than any US federal AI legislation has moved through Congress in the same period and producing precedent the EU AI Act and the UK AI Bill have not yet produced on the same questions. The Andersen plaintiffs converted a January 2023 longshot copyright complaint into an August 2024 federal-court ruling that the case survives to discovery on Lanham Act false-endorsement liability for an artist-style prompting feature, an outcome no observer at filing predicted. The strategy returns concrete, contractually-binding or court-recognised constraints on AI developer behaviour at a pace civil-society legislative advocacy has not matched.

Weaker on its secondary claim — that the wins generalise beyond the unionised and registered tier of the creative workforce. The WGA / SAG-AFTRA MBAs only bind AMPTP signatories; the Andersen class is bounded to artists with US copyright registrations; the Authors Guild class is bounded to authors of registered fiction and nonfiction works; the Concept Art Association federal-policy track represents a self-organised professional community but not the broader visual-artist freelance economy. The estimated several million US freelance illustrators, voice-over artists, gig journalists, self-published authors, and other creative workers outside these institutional perimeters get the strategy's public framing without its protections. The strategy's public-affairs frame ("for everybody in the industry", "for the entire entertainment industry") is structurally over-extended relative to its actual scope of enforcement, and the public-press narrative of "the creator class fights back" is doing some of the work of obscuring this asymmetry.

The deepest structural problem is the doctrinal narrowness of copyright as a labour instrument, addressed at length in the counter_read: above. The two-instrument strategy — collective bargaining where the union exists, copyright class action where it does not — is the strategy's working compromise; whether it can sustain through the 2026 second-round bargaining cycle, the Andersen September 2026 trial, the Authors Guild v. OpenAI fair-use adjudication, and the federal NO FAKES Act / state right-of-publicity legislation track is the open question. The strategy's gains so far are interim; the merits-stage outcomes are still ahead.

Ecology

The strategy is fed by counter-narrative framing through the "existential threat to creative professions" framing that Fran Drescher coined at the 13 July 2023 SAG-AFTRA strike-call press conference and through the "Have I Been Trained?" framing that converted training-data audit into a one-click verification surface individual artists could use without union mediation. The frame did the public-affairs work that made the strike materially viable: it converted what would have been an arcane technical bargaining track into a public-press story the membership could ratify at 99 per cent and 78.33 per cent, and it gave the Andersen and Authors Guild complaints a public-affairs vocabulary the press could carry. The Authors Guild's 18 July 2023 open letter — 15,000+ signatures led by Margaret Atwood, James Patterson, Dan Brown, Nora Roberts, and Suzanne Collins — established the consent-and-compensation framing two months before the September 2023 lawsuit and is the public-organising counterpart of the litigation track.

The strategy uses class-action litigation against private AI as its primary judicial arm. The underlying doctrine is the same — copyright, Lanham Act, DMCA Section 1202, state consumer-protection and right-of-publicity statutes — and the plaintiffs' bar is largely the same (Joseph Saveri Law Firm with Matthew Butterick on Andersen; Lieff Cabraser with Rachel Geman on Authors Guild; a shared SDNY and NDCA technology-docket judicial culture). The distinguishing feature of this strategy is that the class action sits inside a sectoral organising vehicle (the Authors Guild as institutional lead plaintiff, the Concept Art Association as field-defining sectoral organiser) rather than being a free-standing plaintiffs'-bar product. Where the underlying class-action strategy can be run by any plaintiffs' firm against any plausible defendant, this strategy is gated by the existence of a sectoral organising vehicle with institutional standing, member governance, and durable membership renewal — the Authors Guild's 17,000-member roster, the WGA and SAG-AFTRA bargaining apparatus, the Concept Art Association's professional-association infrastructure. Class-action litigation without an organising vehicle behind it is a plaintiffs'-firm portfolio play; class-action litigation inside one is the strategy this verdict is about.

The strategy uses open-letter collective signatory action as its launch vehicle. The Authors Guild's 18 July 2023 open letter preceded and seeded the 20 September 2023 lawsuit; the WGA / SAG-AFTRA Pattern of Demands and public statements ran on letter-form sectoral framings through 2023; the Concept Art Association federal-policy advocacy track and the Human Artistry Campaign entered the public record as multi-signatory letters from sectoral organisations. The letter-form is what gets a sectoral demand into the public-press cycle at launch with the coalition's substantive vocabulary intact, against a defendant whose communications budget would otherwise dominate the framing.

The strategy runs in tandem with organise AI supply-chain workers at the opposite end of the AI supply chain — and the symmetry is structurally instructive. The African Content Moderators Union, the Worker Info Exchange, the Africa Tech Workers Movement and adjacent supply-chain-worker organising axis organises labour at the input end (the content moderation, data annotation, RLHF, gig-dispatch labour that produces and refines training data); the WGA / SAG-AFTRA / Authors Guild / Concept Art Association axis organises labour at the output end (the published, performed, drawn, written corpus the model is trained on). Both are forms of organising creative or precarious labour against the AI supply chain, separated mostly by register: African content moderators organise via union recognition and labour-law instruments in Kenya, South Africa, and Ghana; US writers and actors organise via collective bargaining and copyright class action in Hollywood and the SDNY. The two strategies have not yet substantially collaborated in the public record this corpus tracks; the available coalition between training-data labour and training-corpus labour is one of the corpus's most concrete unbuilt alliances.

The strategy's overlap with strategic litigation against algorithmic state decisions is small but real: both are litigation-anchored strategies running through the US plaintiffs' bar and the federal-court system, and Andersen and Authors Guild share their judicial culture with cases like the ACLU's Williams v. Detroit Police Department. The structural difference is the defendant — private AI developers and their commercial-distribution partners versus state agencies running automated decisions on populations — and the doctrine — copyright, Lanham Act, and state right-of-publicity versus constitutional due process, equal protection, and administrative review. The two strategies are largely separate in adopter, in doctrinal arm, and in the harm they redress; they share a courthouse-floor culture and a plaintiffs'-bar talent pool more than a strategic playbook.

The strategy feeds messages that travel beyond it: the "existential threat to creative professions" frame is the strike cycle's most-cited public-affairs artefact and now travels into adjacent labour-and-AI organising; the "Have I Been Trained?" frame seeded the wider US training-data audit ecosystem and remains the operative public verification surface for individual creators outside the unionised perimeter. The strategy's rhetorical artefacts do work the strategy itself cannot — they reach the under-organised creative workforce the strategy's enforcement perimeter excludes, and they keep that workforce inside the public-press narrative of the strike and the suits even when the contract and the class do not.

The strongest competing strategy on the AI side is the AI developers' licensed-corpus-and-revenue-share counter-offer — OpenAI's news-publisher licensing deals (Associated Press, Axel Springer, News Corp, Le Monde, the Financial Times, Reuters, Vox Media), Anthropic's Bartz v. Anthropic settlement ($1.5 billion administered by the Authors Guild), Google's training-data partnerships with Reddit and Stack Overflow, Adobe Firefly's curated-corpus training. Each is an attempt by the AI developer to convert the strategy's most-organised tier into a paid licensing relationship before the strategy's litigation arm produces an industry-binding precedent. The counter-offer is partly an answer the strategy's adopters wanted (the Authors Guild's Bartz v. Anthropic settlement administration is precisely the consent-and-compensation framing operationalised at scale); partly a tool to fragment the strategy's adopter base by routing around its non-licensable tiers; and partly an instrument to lock in a "willing-licensor" defence against subsequent fair-use adjudication. The strategy's strongest hedge against the counter-offer is the organise AI supply-chain workers axis and the under-organised-creative-workforce constituency that the counter-offer cannot reach — and this is why the unbuilt alliance with supply-chain-worker organising matters: it is the only adjacent organising base whose interests cannot be quieted by a licensing deal.