The Supreme Court says nothing — and that is the ruling.
March 2, 2026. The Supreme Court says nothing. That’s the ruling.
I keep coming back to how quiet it was. No opinion. No dissent worth noting. The Court simply declined to hear Thaler v. Perlmutter, and in that silence it let stand a rule that now governs every AI company telling investors, customers, or a court that “the machine made it.”
Here’s what took me a while to see clearly: the question in Thaler was never whether an AI system can produce something original. Plenty of AI output is original in the ordinary sense of the word. That was never in dispute. What actually mattered was narrower, and more useful: who made the creative choices behind the output, and can that be shown?
Stephen Thaler, a computer scientist, built an AI system he called the Creativity Machine (a modest name, presumably chosen by committee of one). He applied to register a work titled “A Recent Entrance to Paradise,” listing the Creativity Machine as sole author and himself only as claimant, by virtue of owning the machine. He did not claim to have written a prompt, selected among outputs, or edited the result. He told the Copyright Office, in effect, that no human touched the creative decisions at all.
Read that again. That admission is the whole case.
The artifact: the application, not the artwork
If you’re a litigator reading this file, don’t get distracted by the picture. I almost did, the first time. The artifact that decided Thaler v. Perlmutter was the registration record: the field where Thaler described how the work was made, and the total absence of any log, prompt, edit history, or curation record showing a human hand in it.
The Copyright Office denied registration on that basis. Thaler sued under the Administrative Procedure Act, and on August 18, 2023, Judge Beryl Howell granted summary judgment for the Office, writing that the law has never protected works “generated by new forms of technology operating absent any guiding human hand,” and that human authorship is “a bedrock requirement of copyright”. On March 18, 2025, the D.C. Circuit affirmed, holding as a matter of statutory law that only a human being can be an “author” under the Copyright Act, and rejecting Thaler’s argument that the work-made-for-hire doctrine lets a machine fill that role (no word on whether the Creativity Machine was holding out for a corner office and a 401(k)). Thaler asked the Supreme Court to review it; on March 2, 2026, the Court said no.
Three tribunals looked at the same record and reached the same holding: no human hand shown, no copyright.
The case that’s coming is harder
Here’s the part that should worry you more than Thaler’s loss does. Thaler is the easy version of this dispute, because Thaler built his own losing record. He handed the court an admission on a platter. Almost nobody else will make that mistake.
Most AI-assisted work (code, images, contracts, marketing copy, product designs) involves a person prompting, iterating, rejecting outputs, selecting among them, and editing the result. That’s the normal case, the messy human-in-the-loop case, and it’s the one nobody has fully litigated yet. The Copyright Office’s own guidance and the D.C. Circuit’s opinion leave that harder question open: how much human creative control is enough, and how do you prove it happened?
“I wrote the prompts and picked the best version” is a claim about a workflow, and workflows leave a record, or they don’t. A court cannot take a party’s word for how much of a work was human-directed any more than it can take a party’s word for whether an algorithm ran autonomously. Someone has to reconstruct the record: prompt logs, generation parameters, version history, selection and editing steps, timestamps, who had access to what.
The AI vendor can’t credibly do that reconstruction, and neither can the lawyer arguing the case. That job falls to a court-appointed software expert: define the narrow technical question (what did the human actually direct, choose, or change?), list the materials that can answer it (chat logs, model outputs, edit history, version control, export metadata), apply a repeatable method for reconstructing the workflow, state plainly what the record cannot show, and hold up under cross-examination.
Why this needs an expert witness in Israel now
Here’s what I see on the ground: Israeli companies are shipping AI-assisted code, AI-generated marketing content, and AI-drafted deliverables faster than their contracts, cap tables, or employment agreements account for. When a co-founder leaves, when a vendor dispute lands in the Tel Aviv District Court, or when an investor’s diligence team asks “who actually owns this,” the answer increasingly turns on the same question Thaler lost: can you show the human contribution, not just assert it?
Waiting for the dispute to preserve the record is too late. The prompt history, the intermediate drafts, the edit trail: all of that degrades or disappears if nobody thought to keep it. Don’t read Thaler as a ruling that AI content is safe, or that it’s unprotectable. Read it as a ruling that provenance is now evidence, and evidence has to be built, preserved, and tested before a court will believe it.
This is where a source code expert witness earns its place — a document that explains the finding rather than just presenting it.
The above is general information only and does not constitute legal advice. Specific facts of the case cited are drawn from the sources listed.