06-reference/research

ai coloring book copyright resale risk

2026-10-18·research-brief·source: deep-research·by Ray Data Co (deep-research synthesis)
coloring-bookscopyrightcompilationcopyright-officeshopify

The Images Are Unownable, But the Book Is Registrable: Compilation Copyright Is the Asset the Launch Plan Assumes It Doesn't Have

The question

"Beyond confirming QuiverAI's own TOS allow resale, is there a current copyright-law nuance (human-arrangement/compilation copyright, Copyright Office guidance on AI-assisted works) that changes the risk calculus for selling an AI-generated coloring book as a real revenue product on Shopify?"

The 2026-10-06 launch plan treats copyrightability as a settled checkbox inside a vendor-terms question. It is a separate question with a different answer, and the answer is more favorable than the plan assumes.

This is a reading of published Copyright Office materials and a federal docket, not legal advice. Said once; the rest of the brief gives a direct read.

What we already know (from the vault)

What the web says

Convergences and contradictions

Synthesis for RDCO

The backlog note's instinct was right: the TOS checkbox answered "may we sell this" and left "do we own anything" untouched, and those are different questions with different answers. The permission answer was yes. The ownership answer is yes in part, and the part we get is the part that matters commercially. Under Zarya of the Dawn, quoted approvingly in Part 2 §F, a work whose every image was machine-generated was still registered for the human's "selection, coordination, and arrangement of the text and visual elements." That is a coloring book's entire value proposition. Which 25 of the bake-off candidates made it in, what order they run, how image and caption sit on each page, what the cover does - all of it is human work product and all of it is inside the protectable zone. The line art is not, and never will be. The launch plan's accepted-risk line should be amended to say so precisely, because as written it tells the founder's wife she owns nothing, and she will own the thing she actually made.

Separate the three risks the backlog note was right to want separated. (a) Can we be sued for selling it? Low, and low for a reason specific to this product rather than a general reassurance. Infringement needs substantial similarity to a particular protected work, and nativity scenes, angels and shepherds are public-domain iconography with no protected characters anywhere near them. The live training-data litigation is aimed at model developers, not downstream sellers. The real exposure is not the law, it is QuiverAI ToS §14: with no reciprocal IP indemnity, RDCO eats 100% of any claim that does arrive. That was already flagged AMBER on 2026-10-14 and nothing here moves it. The one thing that would move it is subject-matter drift - if any page starts resembling a recognizable commercial character, the analysis changes completely. (b) Can we stop a copycat? This is the question the checkbox did not answer and the answer flipped from "no" to "yes, narrowly." We can stop a wholesale clone of the book. We cannot stop someone generating their own nativity pages in a similar style, and we should not plan as if we could. (c) Does a platform independently restrict it? Shopify: nothing found. KDP: a mandatory private disclosure checkbox that the plan already anticipated, with April-2026 enforcement reportedly tightened, though that enforcement claim rests on trade blogs rather than Amazon's own page.

The actionable changes are cheap and none of them touch the timeline. First, keep a production log of the curation: which designs were chosen from how many candidates, why, in what order, and who made the layout calls. That log is the evidence for the compilation claim and it costs nothing to keep while the book is being built anyway. Second, do the four composition fixes by hand in a vector editor rather than by re-prompting, where that is practical, and note it in the log. Part 2 p. 27 makes hand region-editing an independent route to originality in the individual pages, which would give the book two authorship claims instead of one. Third, register the book after publication, claiming selection/coordination/arrangement plus the human-authored text and any hand-edited elements, and disclaiming the AI-generated images as Part 2 p. 3 requires. Registration is a precondition to suing on a US work (17 U.S.C. §411(a)) and timely registration is what unlocks statutory damages and attorney's fees (§412) - which at a $16.99 price point is the only thing that would ever make a copycat claim worth bringing. Fourth, amend decision #4 in the launch plan so the distinction between the pages and the book is on the record before anyone relies on the current wording.

On whether this needs an attorney before December: not for copyrightability. The doctrine is settled, the Office asked for no new law, the Supreme Court closed the last open door in March, and the registration path is a self-serve online application. Spending legal budget here would buy a restatement of Part 2. The clause that does deserve an attorney is the one the previous brief already named - QuiverAI ToS §14's one-way indemnity - and only if the print run grows past the learning-project scale. Keep the first run small, as planned, and that stays deferred.

Why this is in the vault

It revises decision #4 of [[2026-10-06-shopify-launch-plan]] from "we own nothing, accepted risk" to "we own the compilation, here is how to perfect it," and it adds three concrete pre-launch build steps (production log, hand-editing over re-prompting, post-publication registration) to a book that ships before Christmas. It also closes the copyrightability item that [[2026-10-14-quiverai-tos-commercial-art-license]] explicitly deferred, so the coloring-book legal picture is now complete except for the §14 indemnity question.

Open follow-ups

Related

Sources

Vault:

Primary (retrieved 2026-10-18):

Secondary: