← Blog

Is a Higgsfield video copyrighted? The US, EU and Taiwan on AI video

2026-10-06 · OpenStela

Short answer: a video you make with Higgsfield can contain copyrighted material, but usually not all of it, and not always yours. In the United States and in Taiwan, the official position is the same in outline. What a person created is protected. What the model generated on its own is not. In the EU there is no AI-specific rule yet. The usual originality test applies, and it is built around human choices.

There is a third layer people forget. If you use Genjutsu to put your character into a clip somebody else shot, that footage still belongs to whoever shot it. Generating over it doesn't change that. Below is what each jurisdiction actually says, with links to the official text, and what you can write down today so the human part of your clip is on record. This is general information, not legal advice.

One clip, three layers

Pull a Higgsfield video apart first, because the answer differs for each piece.

1. What you brought in. Footage you shot, character designs you drew, a script, a voice-over, the edit you did afterwards. These are ordinary human work.
2. What the model produced. The motion, lighting, textures and frames the system generated from your prompt and inputs.
3. What someone else made. A movie scene, a dance video or a stranger's TikTok that you used as the source clip. Higgsfield's announcement of Genjutsu (August 31, 2026) says it works on "either real footage or generated video", from one input clip of up to 30 seconds.

The next three sections cover layers 1 and 2. Layer 3 gets its own section, because it is the one that gets creators into trouble.

United States: only the human part is protected

The U.S. Copyright Office set out its position in Copyright and Artificial Intelligence, Part 2: Copyrightability, published January 29, 2025. Its summary of conclusions is short (p. iii). Three lines matter for video:

• "Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements."
• "Based on the functioning of current generally available technology, prompts do not alone provide sufficient control."
• "Human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs, as well as the creative selection, coordination, or arrangement of material in the outputs, or creative modifications of the outputs."

For Higgsfield users, the most useful part is the section on what the report calls expressive inputs. If you feed in your own copyrightable work and "that work is perceptible in the output, they will be the author of at least that portion of the output" (p. 24). The protection covers "the perceptible human expression", not the AI-generated elements standing alone. A dance you choreographed and filmed, carried through Motion Transfer, is a different situation from typing a sentence and accepting the result.

The report also says that putting AI material inside a larger human work doesn't sink the whole: "a film that includes AI-generated special effects or background artwork is copyrightable, even if the AI effects and artwork separately are not" (p. 27). Its conclusion (p. 41) puts the prompt question plainly: "Prompts alone, however, at this stage are unlikely to satisfy those requirements." The Office also says this is decided case by case. Nothing here turns a particular clip into a yes or a no.

European Union: the originality test, plus a labelling rule

There is no EU statute that says who owns AI output. The test that applies to any work comes from the Court of Justice. In Painer (C-145/10), decided December 1, 2011, the Court said "an intellectual creation is an author's own if it reflects the author's personality" (para. 88). That is the case "if the author was able to express his creative abilities in the production of the work by making free and creative choices" (para. 89). The case was about a portrait photograph, not AI. How national courts apply it to generated footage is for them, and we have no ruling on AI video to cite.

What the EU has added is a disclosure rule. The AI Act, Regulation (EU) 2024/1689 (June 13, 2024), defines a "deep fake" as AI-generated or manipulated image, audio or video content that "resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful" (Art. 3(60)). Article 50(4) says deployers of a system that produces deep fakes "shall disclose that the content has been artificially generated or manipulated". For "evidently artistic, creative, satirical, fictional" work, that is limited to disclosing the AI content "in an appropriate manner that does not hamper the display or enjoyment of the work". A "deployer" excludes use "in the course of a personal non-professional activity" (Art. 3(4)). Article 113 of the original text sets 2 August 2026 as the general application date. Read the current consolidated version on EUR-Lex before relying on any date.

This rule is about labelling, not ownership. But if you publish realistic Genjutsu footage to EU audiences as part of a business, it is a separate obligation to check.

Taiwan: it depends on human creative input

Taiwan's Intellectual Property Office answered this question directly in an interpretation letter, 電子郵件1141017b, dated October 17, 2025. Read with Articles 3 and 10 of the Copyright Act, it says:

• Whether an AI-generated work is protected depends on whether a human put actual creative input into the process.
• If AI is used as an assisting tool (the letter's example is drawing software) and there is real human creative input, the result can be protected. The person who actually created it generally holds the copyright.
• If the content was produced entirely by the AI's own computation with no human creative input, it is not protected.
• Because copyright is a private right, disputes over whether there was enough human input are for the courts to decide on the facts of each case.

The question that prompted the letter was about an AI assistant for drawing characters and virtual people. So the letter fits character creators closely. TIPO also held a public hearing on a draft generative AI copyright guideline on July 20, 2026, according to its own monthly summary. Check TIPO's site for whether a final version has been published since.

When the footage isn't yours

This is where Genjutsu differs from text-to-video. Motion Transfer and Object Swap start from a real clip. If that clip is your own, it is layer 1, your work. If it is a film scene, a music video or someone else's post, it is layer 3, and every system above protects it regardless of what you generate on top.

• US. Under 17 U.S.C. § 103, protection for a work that uses preexisting copyrighted material "does not extend to any part of the work in which such material has been used unlawfully" (§ 103(a)). Even where use is lawful, copyright in a derivative work "extends only to the material contributed by the author of such work" and "does not imply any exclusive right in the preexisting material" (§ 103(b)).
• Taiwan. The Copyright Act gives the author the exclusive right to adapt a work into a derivative work (Art. 28). A derivative work is protected on its own, but that protection "對原著作之著作權不生影響", meaning it has no effect on the copyright in the original (Art. 6).

Swapping your character into a film clip doesn't make the clip yours, or give you permission to use it. Whether a particular use is fair use (US) or 合理使用 (Taiwan) is a fact-specific question for a lawyer, not something a generator decides.

What Higgsfield's terms add

Copyright law decides what is protectable. Your contract with the tool decides what you and the platform may do with inputs and outputs. That contract is Higgsfield's Terms of Use. When we checked on October 6, 2026, the page showed a "Last updated" date of July 26, 2026. It has an "Ownership" section with a subsection titled "User Inputs and Outputs", and a separate "AI Disclosure" subsection. Read those sections for your account and plan, in the current version. We don't summarise them; terms change.

Writing down the human part

Across all three systems, the deciding question is the same: what did a person do, and was it theirs to use? That is a matter of evidence, and evidence is easiest to make on the day you finish.

When you register a work on OpenStela, the record asks for exactly that:

1. Open the works page and register the file. If it is on Higgsfield, you can paste the file's address and we download the original. The fingerprint is the SHA-256 of the video itself.
2. Tick This work was made with AI tools and list them. Once that box is ticked, Your contribution is required. Write acts, not adjectives: "Shot the source clip myself; drew the character sheet; Motion Transfer preset; cut and scored in Resolve."
3. Under Rights, choose My own work or I hold a licence. If the source clip came from someone else and you have permission, that is the honest box.
4. Register. The record is fingerprinted at once and anchored to a public chain in that night's batch. Nobody can add to it later, including you.

What you get is a dated, verifiable record: this file existed on this date, under this name, with this description of what a person contributed. It is evidence if a dispute comes up. It does not decide whether you hold copyright, it doesn't make someone else's footage yours, and the contribution text is your own statement, not something we check. Made a character with Genjutsu? Do this before it goes viral walks through registering the character first and binding each clip to it. Can AI art be copyrighted? covers what to write in the contribution field.

Registering works is free. Uploads share 300 MB of space per account, and link-type works take almost none (see pricing).

This is general information, not legal advice. If a real dispute comes up, a lawyer where you live is the right person to weigh it.