Building your AI product after you raise: a founder's guide / Build the product
Who owns the code when a development partner or an AI writes it
A startup has to own the code it sells, and investors and buyers check that it does. Under US law, copyright first belongs to the author, an employee's work within the job belongs to the employer, and a transfer of copyright from a contractor must be in writing and signed. The US Copyright Office said in January 2025 that purely AI-generated material is not protected by copyright, while creative selection, arrangement and changes to AI output by people can be. This page explains each rule with its source, the Thaler court decision, the UK rule, and what to put in writing with a development partner.
Published October 26, 2026. Editorial.
Key takeaways
- Under 17 U.S.C. section 201, copyright first belongs to the author. For a work made for hire, the employer or the person who commissioned it counts as the author.
- A commissioned work is a work made for hire only in nine listed categories and with a signed written agreement. The list does not name computer programs, so ask a lawyer to draft the contract.
- Under section 204, a transfer of copyright is valid only if it is in writing and signed by the owner of the rights.
- The Copyright Office says purely AI-generated material is not protected, prompts alone do not give enough control, and creative selection, arrangement and changes to AI output by people can be protected.
- In March 2026 the Supreme Court denied review in Thaler v. Perlmutter, leaving in place the ruling that the Copyright Act requires a human author.
Investors in the next round, and any later buyer of the company, will check that the company owns its code. Two facts make that question harder for an AI startup. Some of the code may be written by people outside the company, such as a development firm or a freelancer. And some of it may be written by an AI model. This page explains what US copyright law and the US Copyright Office say about each case, with the sources. It is general information, and a lawyer should draft and review your contracts. It is part of the guide on building your AI product after you raise.
Rule one: copyright starts with the author
Copyright is the legal right to copy and sell a creative work, such as written code. The starting rule is in 17 U.S.C. section 201(a): "Copyright in a work protected under this title vests initially in the author or authors of the work" [5]. "Vests initially" means the right belongs to the author first.
There is one important exception, the work made for hire. Section 201(b) says: "In the case of a work made for hire, the employer or other person for whom the work was prepared is considered the author for purposes of this title" [5]. So for a work made for hire, the company is the author from the start, and no transfer is needed.
Rule two: employees and contractors are treated differently
Section 101 of the same title defines a work made for hire in two ways [4]. The first is "a work prepared by an employee within the scope of his or her employment". Code that an employee writes as part of the job is a work made for hire, and the company owns it.
The second covers commissioned work, which is work ordered from someone outside the company. It counts as a work made for hire only if it falls into one of nine listed categories, and only "if the parties expressly agree in a written instrument signed by them that the work shall be considered a work made for hire" [4]. The nine categories are a contribution to a collective work, part of a motion picture or other audiovisual work, a translation, a supplementary work, a compilation, an instructional text, a test, answer material for a test, and an atlas. Computer programs are not named in that list.
The Copyright Office's Circular 30, revised in August 2024, makes the test strict. It says a commissioned work is a work made for hire only "if it satisfies all of the following four criteria", and "If a work fails to satisfy any of these requirements, it is not a work made for hire" [7]. It also says the question "is determined by facts in existence at the time the work is created" [7]. Sign the agreement before the work starts, and ask a lawyer about code written before it.
Rule three: a transfer must be in writing and signed
When the company is not the author, it needs the author to transfer the copyright. Section 204(a) sets the form: "A transfer of copyright ownership, other than by operation of law, is not valid unless an instrument of conveyance, or a note or memorandum of the transfer, is in writing and signed by the owner of the rights conveyed or such owner's duly authorized agent" [6]. An instrument of conveyance is a document that transfers ownership, often called an assignment.
For code from a contractor or a development firm, you therefore need a written, signed assignment of the copyright to your company. A lawyer should draft the agreement, and it should cover code already written as well as code to come.
| Who wrote the code | Who owns the copyright at first | What the company needs |
|---|---|---|
| An employee, within the job | The company, as a work made for hire [4] [5] | An employment agreement that says so clearly |
| A contractor or a development firm | The contractor, as author [5] | A written, signed assignment to the company [6] |
| A founder before the company existed | The founder, as author [5] | A written, signed assignment to the company [6] |
| An AI model, with no human creative control | No one: purely AI-generated material is not protected [1] | Creative selection, arrangement or changes by people, which the law can protect [1] |
| A person working with AI tools | The person, for the human expression in it [1] | The same employment or assignment paperwork as above |
What the Copyright Office says about AI-written material
The US Copyright Office published Part 2 of its report on copyright and artificial intelligence in January 2025. Its main conclusions, in its own words [1]:
- "Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements."
- "The use of AI tools to assist rather than stand in for human creativity does not affect the availability of copyright protection for the output."
- "Copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material."
- "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."
- "Whether human contributions to AI-generated outputs are sufficient to constitute authorship must be analyzed on a case-by-case basis."
The Office concluded that these questions "can be resolved pursuant to existing law, without the need for legislative change" [1]. It also said the answer could change: "There may come a time when prompts can sufficiently control expressive elements in AI-generated outputs to reflect human authorship" [1].
For a startup, the report means this. Code that an AI model writes from a prompt, with no human creative choices, is purely AI-generated material, which the Office says copyright does not protect. Code in which people make the creative choices, by selecting, arranging and changing what the model produced, or by writing parts themselves, can carry human expression that copyright protects, and the Office says each case is decided on its facts. Keep records that show those human choices: the specification, the design decisions, the review comments and the changes people made.
The Thaler decision
The courts have applied the same rule. In Thaler v. Perlmutter, decided on March 18, 2025, the US Court of Appeals for the D.C. Circuit held that an AI system called the Creativity Machine "cannot be the recognized author of a copyrighted work because the Copyright Act of 1976 requires all eligible work to be authored in the first instance by a human being" [2]. The court added that "the human authorship requirement does not prohibit copyrighting work that was made by or with the assistance of artificial intelligence" [2].
Stephen Thaler asked the Supreme Court to review the case. Its docket for case 25-449 records, on March 2, 2026: "Petition DENIED" [3]. The appeals court ruling stands.
Outside the United States
Other countries have their own rules. The UK's Copyright, Designs and Patents Act 1988 has a rule for computer-generated works: "the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken" [9]. The Copyright Office's report summarises two other positions [1]. It says EU member states, answering a 2024 questionnaire, shared the view that AI-generated content may be eligible for copyright "only if the human input in [the] creative process was significant." It says Japan's guidelines of May 2024 decide the question case by case. This guide read only the Copyright Office's summary of those two, and did not open the EU or Japanese documents. If you sell in those countries, ask a local lawyer.
The model provider's terms
Providers also write terms about who owns the output. Anthropic's commercial terms, effective June 17, 2025, say: "Subject to Customer's compliance with these Terms, Anthropic hereby assigns to Customer its right, title and interest (if any) in and to Outputs" [8]. The words "if any" matter. The terms pass to you whatever rights Anthropic has, and they cannot create a copyright that the law does not give. This guide could not open OpenAI's services agreement, so it does not describe OpenAI's terms on outputs.
Open-source code is part of ownership too
Your codebase also contains code that nobody at the company wrote: open-source libraries, published under licences with conditions. Lawyers at Morgan Lewis, writing in June 2026 about due diligence for a company sale, list "a review of the company's practices relating to the use of software licensed under copyleft licenses, such as the GNU General Public License (GPL), Affero General Public License (AGPL), Lesser GPL (LGPL), and similar licenses" among the checks [10]. A copyleft licence can require a company that distributes the code to publish its own changes under the same licence. Run a licence scan on all the code, whoever or whatever wrote it, as well as keeping your assignments. The page code provenance and licence exposure in AI-generated code covers what investors check.
What to put in writing with a development partner
Ask a lawyer to draft the agreement. The points to raise with them:
- An assignment that transfers to your company, from the day it is signed, the copyright in all code, documents and tests the partner writes for you, signed by the partner.
- The same assignment for any freelancer or subcontractor the partner uses.
- A statement of how the partner uses AI tools, and records of the human review and changes, so that you can show the human expression in the code.
- A list of open-source components with their licences, delivered with the code.
- Delivery of the eval suite, the specification and the prompts as well as the code, so that you own everything needed to change the product later.
The page what investors expect before the next round shows where these records go in a data room.
Where Reveneau fits
Reveneau is an AI software development consultancy with offices in San Francisco and Vancouver. Reveneau says AI writes all of its code and every change must pass an eval suite written from the specification, which Jev, TypeSafe AI's decision model, grades. It says it takes responsibility for the whole project through production and after release. Ask any partner, Reveneau included, to put the assignment of copyright, the handover of the eval suite and the open-source list into the contract. To discuss a build, see AI development at Reveneau or contact Reveneau. The page an AI development partner for a startup compares the kinds of firm a startup can hire.
Common questions
Who owns code written by an employee?
Code written by an employee within the scope of the job is a work made for hire under 17 U.S.C. section 101, and section 201(b) says the employer is then considered the author. The company owns it from the start. An employment agreement that states this, and covers inventions and code, makes the position clear to investors and buyers.
Who owns code written by a contractor or a development firm?
Code written by a contractor or a development firm belongs at first to its author, under 17 U.S.C. section 201(a), unless it qualifies as a work made for hire. Section 204 says a transfer of copyright is valid only if it is in writing and signed by the owner. Get a signed assignment to your company, drafted by a lawyer, covering code already written and code to come.
Is code written by an AI protected by copyright?
Purely AI-generated code is not protected by copyright in the United States, according to the US Copyright Office's January 2025 report, which says copyright does not extend to material where there is insufficient human control over the expressive elements. Code in which people creatively select, arrange or change the AI's output, or write parts themselves, can carry human expression that copyright protects, decided case by case.
Are prompts enough to make me the author of AI output?
Prompts alone are not enough, according to the US Copyright Office, which wrote in January 2025 that, based on current generally available technology, prompts do not alone provide sufficient control. It added that a time may come when prompts can control the expressive elements of AI output. Each case is analysed on its own facts.
What did the court decide in Thaler v. Perlmutter?
The US Court of Appeals for the D.C. Circuit decided on March 18, 2025 that an AI system cannot be the author of a copyrighted work, because the Copyright Act of 1976 requires a human author. The court added that the rule does not bar copyright in work made with the help of AI. The Supreme Court denied review on March 2, 2026.
What is a work made for hire?
A work made for hire is a work whose author, under US copyright law, is the employer or the person who commissioned it. Section 101 covers an employee's work within the job, and a commissioned work in nine listed categories with a signed written agreement. Circular 30 says a commissioned work that fails any of the four criteria is not a work made for hire.
Can a contract signed after the code was written make it a work made for hire?
The Copyright Office's Circular 30, revised in August 2024, says whether a work is a work made for hire is determined by facts in existence at the time the work is created. Sign the agreement before the work starts. For code already written, ask a lawyer about a written and signed transfer under 17 U.S.C. section 204.
Does Anthropic give me ownership of Claude's output?
Anthropic's commercial terms, effective June 17, 2025, say Anthropic assigns to the customer its right, title and interest, if any, in outputs. The words if any matter: the terms pass on only the rights Anthropic has, and they cannot create a copyright that the law does not give. The same terms say the customer owns its outputs.
Is AI-generated work treated the same way in the UK?
The UK has its own rule. Section 9(3) of the Copyright, Designs and Patents Act 1988 says that for a computer-generated work, the author is the person who made the arrangements necessary for its creation. That differs from the US position in the Copyright Office's January 2025 report. A company selling in the UK should ask a UK lawyer how the rule applies.
What should a development contract say about code ownership?
A development contract should assign to your company, in writing and signed, the copyright in all code, documents and tests the partner writes, because 17 U.S.C. section 204 requires a signed writing for any transfer. It should cover subcontractors, list open-source components with their licences, and hand over the eval suite and specification. A lawyer should draft it.
Why does open-source code matter for ownership?
Open-source code in your product comes with licence conditions that you must meet. Lawyers at Morgan Lewis, writing in June 2026 about due diligence for a company sale, list a review of copyleft licences, such as the GPL, AGPL and LGPL, among the checks. A copyleft licence can require a company that distributes the code to publish its own changes, so keep a list of components and run a licence scan.
References
- [1] US Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability (January 2025): purely AI-generated material is not protected; AI assistance does not remove protection; prompts alone do not give enough control; human selection, arrangement and changes can be protected; case by case; the summaries of the EU and Japanese positions.
- [2] US Court of Appeals for the D.C. Circuit, Thaler v. Perlmutter, No. 23-5233 (decided 18 March 2025): the Copyright Act requires a human author; the rule does not bar works made with AI's help.
- [3] Supreme Court of the United States, Docket No. 25-449, Stephen Thaler v. Shira Perlmutter: petition denied on 2 March 2026.
- [4] 17 U.S. Code section 101, Definitions, Legal Information Institute, Cornell Law School: work made for hire by an employee, or a commissioned work in nine categories with a signed written agreement.
- [5] 17 U.S. Code section 201, Ownership of copyright, Legal Information Institute, Cornell Law School: copyright vests initially in the author; for a work made for hire the employer is the author.
- [6] 17 U.S. Code section 204, Execution of transfers of copyright ownership, Legal Information Institute, Cornell Law School: a transfer must be in writing and signed.
- [7] US Copyright Office, Circular 30, Works Made for Hire (revised August 2024): all four criteria for a commissioned work; facts at the time the work is created decide.
- [8] Anthropic, Commercial Terms of Service (effective 17 June 2025): Anthropic assigns to the customer its right, title and interest, if any, in outputs.
- [9] UK Copyright, Designs and Patents Act 1988, section 9, legislation.gov.uk: for a computer-generated work, the author is the person who made the arrangements necessary for its creation.
- [10] Morgan, Lewis & Bockius LLP, Open-Source Software: Common Areas of Inquiry in M&A Due Diligence (5 June 2026): review of copyleft licences such as GPL, AGPL and LGPL.
Related reading
Why code ownership still matters when agents write it
An agent can write a service in an afternoon, but somebody still has to be the person who gets the alert at two in the morning, and that job does not disappear just because a human did not type the code.
What changes in a contract when AI writes the code
When AI writes the code, three clauses need new wording and the rest of the contract can stay exactly as it was.
How to negotiate a software contract you can actually verify
Most build contracts describe effort, timeline, and payment, and leave the one hard question unanswered: on what evidence do you agree the thing is finished?
How to prepare for technical due diligence before a raise or sale
Technical due diligence is where a deal can quietly fail. Here is what investors' technical reviewers actually look at, how to prepare before they do, and the warning signs that worry them.
More in Build the product
API or your own model: how a startup should build the AI part
A startup can put AI into its product in three ways: call a model that a provider hosts, through its API; fine-tune a hosted model on its own examples; or run an open-weight model on servers it controls. As read on 26 October 2026, OpenAI is winding down its self-serve fine-tuning and has closed it to new users, and Google's Gemini API has no model that supports fine-tuning. Reveneau's view is that a startup should start with a hosted model, clear instructions and an eval suite, and change approach only when the evals show a gap. This page sets out each option with the providers' own terms.
What an AI feature costs to run, per user, with October 2026 prices
An AI feature costs money every time a user uses it, because the model provider charges for each piece of text sent and received. This page builds one invented example, a support assistant used 40 times a month per user, and prices it with the rates OpenAI, Anthropic and Google published on 26 October 2026. On the mid-priced models the example costs $0.44 per user per month, and on the smallest models $0.022. The page then shows how prompt caching and batch requests change the figure, which charges add to it, which prices are scheduled to change, and the spending limits each provider sets.
Choosing a model provider: what OpenAI, Anthropic and Google keep and use
Before customer data goes into a model, read what the provider keeps, for how long, where, and whether it trains on it. As read on 26 October 2026, OpenAI says API data is not used for training unless the customer opts in, and keeps abuse monitoring logs for up to 30 days. Anthropic's terms say it may not train on customer content, and it deletes API inputs and outputs within 30 days with stated exceptions. Google uses prompts sent to its free Gemini API services to improve its products and keeps abuse monitoring data for 55 days. This page sets out each provider's terms side by side.