Practises
3 links
Graph · Organisation
01 · In focus
The structured facts the source records about Fairly Trained, the count of declared adjacencies in the corpus, and the federation map zoomed on this node and its neighbours.
organisation
↑9 declared connections
02 · Connections
Split by direction. Direct links are the ones Fairly Trained’s source record names; inferred backlinks are records elsewhere in the corpus that point at this entity. Some records appear in both because the corpus names them from both sides — those rows carry a note.
6 links
Links named in this entity's structured fields.
3 links
3 links
3 links
Other records that name this entity.
03 · Background
Body prose as it appears in movement-graph’s published markdown for this entity. Links to other corpus entities resolve to their graph page; links to deeper repo paths are kept as text so the page does not invent a route.
Fairly Trained is an international nonprofit founded in January 2024 to certify generative AI models that train only on licensed data — a Fair Trade–analogous mark for the creator-consent movement around AI training practices. The organization was founded by Ed Newton-Rex, a British composer and technologist who resigned from Stability AI in November 2023 after publicly disagreeing with the industry position that training AI on copyrighted material without consent is permissible under fair use.
Newton-Rex joined Stability AI in 2022 to lead its audio team and oversaw the launch of Stable Audio, named one of TIME Magazine's best inventions of 2023. He resigned in November 2023 after concluding that the company's stance — and the prevailing AI industry position — that non-consensual training on copyrighted material constitutes fair use was ethically untenable and structurally threatening to creative workers' livelihoods. Rather than joining another AI company, Newton-Rex founded Fairly Trained to build certification infrastructure for the consent-based alternative.
The organization's core framing holds that a fundamental divide is opening among generative AI companies: those who obtain consent from data providers and those who do not. Fairly Trained's certification mark is designed to make that divide legible to consumers, partners, and the creative industry so that market mechanisms can reward consent-respecting developers — an application of the Fair Trade model's insight that consumers will differentially reward certified ethical practice when the signal is credible and visible at the point of purchase.
Fairly Trained's primary instrument is the Licensed Model (L) certification, awarded to any generative AI model that uses no copyrighted work without a license. To obtain certification, companies must disclose their training data sources and the licenses under which they were obtained. The certification is binary — a model either meets the consent standard or it does not; there is no scoring tier or partial compliance.
At launch in January 2024, nine generative AI companies received certification, all operating in audio and image: Beatoven.AI, Boomy, BRIA AI, Endel, LifeScore, Rightsify, Somms.ai, Soundful, and Tuney. By March 2024, the program expanded to certify KL3M — its first large language model, from 273 Ventures — and Voicemod, its first certified speech and singing AI, alongside additional audio companies. As of 2025, the program had certified roughly 20 developers and models spanning audio, image, text, and speech modalities, with applications continuing to be accepted.
Fairly Trained launched with institutional support from Universal Music Group, the Association of American Publishers (AAP), the Association of Independent Music Publishers, and Concord. By March 2024, The Authors Guild, SAG-AFTRA, and the Music Managers Forum had joined as supporters, bringing direct trade-union and professional-association weight to the certification's legitimacy signal. The advisory committee includes AAP president and CEO Maria Pallante, Tom Gruber (co-inventor of Apple's Siri), and composer Max Richter.
The breadth of support — spanning book publishing, recorded music, music management, and performers' union representation — reflects Fairly Trained's positioning as a cross-industry consent-infrastructure player rather than a single-sector advocate. The supporting institutions collectively represent millions of creative workers whose livelihoods depend on licensing norms in AI training.
Beyond certification, Fairly Trained and its founder organized two major creator-class mobilization campaigns. Newton-Rex spearheaded an open letter condemning unlicensed AI training that gathered over 50,000 signatures, including Thom Yorke of Radiohead, Julianne Moore, James Patterson, and Kevin Bacon — one of the largest coordinated creator-class public statements on AI training data practices.
In February 2025, Newton-Rex organized a silent album protest — Is This What We Want? — targeting proposed UK copyright law changes that would have allowed AI developers to train models on any lawfully-accessed material without consent. The album featured more than 1,000 musicians — including Kate Bush, Annie Lennox, and Cat Stevens — who recorded 12 tracks of silence in studios and empty spaces across the UK, combining to carry the message: "The British government must not legalise music theft to benefit AI companies." The protest contributed to sustained public and parliamentary pressure on the UK government's AI copyright consultation.
Newton-Rex was named to the TIME100 AI list for 2025, placing Fairly Trained's work among the year's most consequential interventions in AI governance.
04 · Sources
7 sources listed from the pinned corpus. Links are shown only when the source URL is a valid HTTP(S) address.
Official website — certification program description, certified models list, supporters and advisors
January 2024 launch blog post — Licensed Model certification definition; nine initial certified companies; four inaugural industry supporters including Universal Music Group and AAP
Founder blog — Ed Newton-Rex founding rationale; resignation from Stability AI November 2023; consent-vs-fair-use framing; nonprofit structure
VentureBeat January 2024 — launch coverage; certification mechanics; advisory committee including Tom Gruber (Siri co-inventor) and Maria Pallante (AAP CEO)
March 2024 expansion — KL3M (first certified LLM from 273 Ventures); Voicemod (first certified speech and singing AI); new supporters Authors Guild, SAG-AFTRA, Music Managers Forum
TIME100 AI 2025 — Newton-Rex listed; open letter 50,000+ signatures; silent album 1,000+ musicians protesting UK copyright softening; ~20 certified developers by 2025
Al Jazeera February 2025 — silent album protest; 1,000+ musicians including Kate Bush, Annie Lennox, Cat Stevens; 12 tracks combining to form the message against UK copyright changes enabling AI training on any lawfully-accessed material
Source: entities/organizations/org-fairly-trained.md — movement-graph pin 5edfc3b.