Collaborator, or threat to human art? Both sides, steel-manned, with every claim traced back to a named source. Undated on purpose — this issue gets refreshed as the argument moves.
Eight parts, one cover. Same frame every issue in The Debate; only the words and the two filled grid cells change.
For years the question was whether a machine could make art at all. That is no longer where the heat is. By 2026 the argument is narrower and more practical: did the people whose work trained these models agree to it, and do they get paid when the output competes with them?
That shift matters because it is answerable. "Is it art?" is a question for critics. "Was there consent, and is there payment?" is a question for courts, contracts and licences — and all three started moving in 2025 and 2026. This issue lays out the strongest version of each side, then shows where the line is actually being drawn.
Each side, put as well as its own advocates would put it. Sources named. No figure here is offered as settled fact; each is a claim by a named author, linked in full at the end.
UNESCO's 2026 monitoring report, covering more than 120 countries, projects generative AI could cut music-creator revenues by about 24% and audiovisual income by about 21% by 2028.
— UN News, reporting UNESCO, Feb 2026
A January 2026 survey found roughly one in ten Japanese manga artists and illustrators reported income falls tied to AI; 2.7% said they lost more than half.
— The Japan Times, Jan 2026
Brian Merchant reports well over half of surveyed artists say they have lost income to image generators, and 36% of illustrators see fewer commission inquiries.
— Brian Merchant, Blood in the Machine
A court found training on books can be fair use, but holding pirated copies is not; Anthropic settled with authors for about $1.5bn. Scott Turow and two publishers later sued Meta on similar grounds.
— NPR / Authors Guild, 2025–2026
The US Copyright Office found that using AI as a tool does not by itself remove copyright — a human's creative contribution is still protectable, judged case by case. Prompts alone are not enough.
— US Copyright Office, Report Part 2, Jan 2025
Stanford HAI scholars are building open tools so artists can direct models the way they sketch then refine. Their framing: the gap is control, not capability.
— Stanford Report, Mar 2026
An Artlist creator survey found 87% of creators now use AI — 37% for ideation, 26% for faster editing — and most describe it as a co-pilot for the dull parts, not a replacement.
— Artlist survey, via TechCrunch, 2026
Some artists set their own terms rather than refuse the tool: Grimes licenses her voice model on an opt-in basis, expanded in 2026 with tiered payment for different uses.
— Soundverse, 2026 list of pro-AI artists
House rule: any percentage about jobs or income is attributed to its author and study. We reference claims; we do not manufacture certainty. No one knows the true long-run figure.
Both sides are circling the same answer from different directions: artists need a say in whether their work trains a model, and a share when it does. The clearest test of that idea so far comes from music.
In October 2025 Universal Music settled its lawsuit with the AI music startup Udio and agreed an opt-in licence — the artist chooses in, rather than having to chase the model down to opt out. The two are building a joint platform for 2026 where people remix licensed catalogue inside a "walled garden", so nothing leaves the system unlicensed. Warner Music struck a deal with Suno soon after, and Suno raised $250m, with NVIDIA's venture arm among the backers.
It is not a finished model. Reports say Universal's talks with Suno later stalled with no clear path. But the shape of a settlement is now visible: consent by default, payment by licence, and control over what the tool is allowed to produce. That is a long way from "machine versus human."
Worth naming the gap: label deals protect catalogue owners and signed artists. They do less for the freelance illustrator or the unsigned musician whose work was already scraped. A licence between large companies is not the same as a wage for a working artist.
Three places, right now, are deciding what the practical answer looks like:
US rulings have split the difference: training on lawfully obtained work has been treated as transformative and potentially fair use, while using pirated copies has not. The Anthropic settlement and the Turow–Meta suit are testing where that line sits in practice.
The music deals are the first attempt to price consent rather than litigate it. If they hold, expect the same opt-in, pay-per-use shape to spread to image and text.
The quieter answer is in how the work gets made. The Copyright Office and the Stanford researchers point the same way: the more a person shapes, directs and edits the output, the more it is theirs — legally and creatively. Used as a tool under a human hand, AI looks less like a threat and more like the camera did to painting: unsettling, then absorbed.
Every claim above, traced to its author. Read the originals; argue with them, not with us.
The Debate — No. 06 · AI and Creativity · Craig Stanley Studio · CS·PRESS · Undated by design, refreshed as the argument moves · One red, used once · All colour pairings meet WCAG 2.1 AA.
Nine debates, each titled and never dated, each refreshed in rotation as the argument moves. Plain English. Both sides put at their strongest. Every claim about the world attributed to its author with a link. No hype, no banned words, no manufactured certainty.
AI as a force for good · AI in work · AI in education · AI and young people · AI and ethics · AI and truth · AI and the environment · AI and power.