Developing an AI Product with an External CTO: Who Owns the Technology?
Key takeaways
- Default rule: In the UK, EU and US, a person who is not an employee generally owns the copyright in the code and other materials they create, even if the company has paid for the work or issued them shares.
- Written assignment: Ownership passes to the company only through a written, signed assignment. In the US, a “work made for hire” clause will not usually be enough for software created by a consultant.
- AI components: An AI product includes more than code: models, weights, training data, prompts and pipelines all need to be covered, and some are best protected as confidential information rather than copyright.
- Fix it early: Gaps are usually identified in investor or buyer due diligence, when the external CTO’s bargaining position is strongest. Agreements are best put in place before development starts.
Many early-stage businesses build their first AI product without an in-house technical team. A founder with a commercial idea may engage a fractional or interim CTO, a freelance machine learning engineer, or a technical adviser who is promised equity in place of a full salary. The arrangement is often informal at the outset, documented by an exchange of emails, a short consultancy letter or nothing at all.
This creates a question that tends to surface at the worst possible moment, usually during a funding round, an acquisition or a falling-out between the founders: does the company actually own the technology the external CTO built? This article considers the position in the UK, with comparison to the EU and US, and the steps a business can take to secure ownership.
Why does it matter who owns technology?
For an AI business, the technology is often the main asset. If the company cannot show that it owns its code and models, it may be unable to stop the external CTO from reusing them for a competitor, unable to enforce its rights against third parties, and unable to give the intellectual property warranties that investors and buyers expect. A gap in ownership can delay a transaction, reduce its value, or give the external CTO considerable leverage to negotiate further payment or equity before signing the documents needed to close it.
Who owns the code an external CTO writes in the UK?
The default position: the author owns it
Under section 11 of the Copyright, Designs and Patents Act 1988 (CDPA), the author of a work is its first owner. The main exception, in section 11(2), applies only where a work is made Computer programs, and the preparatory design material for them, are protected as literary works.
An external CTO engaged as a consultant or contractor is not an employee for this purpose. In the absence of an agreement to the contrary, the CTO will therefore own the copyright in the code they write, even though the company has paid for it. Being appointed as a director, or holding shares, does not by itself make someone an employee or change that position.
Assignments must be in writing
Copyright can be transferred by assignment, but under section 90(3) of the CDPA an assignment is not effective unless it is in writing and signed by or on behalf of the assignor. Section 91 allows future copyright to be assigned in advance, so that it vests in the company automatically as the work is created. This is why an assignment is best signed at the start of the engagement rather than on delivery or completion.
Implied terms are a limited safety net
Where there is no written assignment, the courts may imply a term into the engagement. In Robin Ray v Classic FM [1998] FSR 622, approved by the Court of Appeal in R Griggs Group Ltd v Evans [2005] EWCA Civ 11, the court explained that the mere fact that a contractor has been commissioned does not entitle the client to the copyright, and that any implied term will be the minimum necessary to give business efficacy to the arrangement. Usually that is a licence to use the work for the purpose for which it was commissioned. An assignment, or a beneficial interest in the copyright, will be implied only where the client needs the right to exclude the contractor and others from using the work.
A company relying on an implied licence or an equitable claim is in a much weaker position than one with a signed assignment, particularly in due diligence, where investors will expect to see a clear documentary chain of title.
Inventions, databases and know-how
Copyright is not the only right involved. Under the Patents Act 1977, the right to apply for and obtain a patent generally starts with the inventor, subject to statutory rules and prior enforceable agreements. Under section 39, inventions belong to an employer if made by an employee during normal or assigned duties where an invention was reasonably expected, or if the employee had a special obligation to further the interests of their employer. Therefore, patentable inventions made by an external CTO also need to be assigned. Database right in a dataset belongs to its maker which means the person who takes the initiative in obtaining, verifying or presenting the contents and assumes the investment riskwith the employer treated as the maker only where the database is made by an employee in the course of employment. Confidential information and trade secrets, which may be the most important protection for model weights, training methods and data pipelines, depend on appropriate confidentiality obligations being in place.
What parts of an AI product need to be covered?
Consultancy agreements are often drafted with conventional software in mind. An AI product typically includes a wider range of materials, and the agreement should make clear that each of them belongs to the company:
- Source code and architecture: application code, model architecture, training scripts, data pipelines, evaluation tools, and design documents.
- Models and weights: trained and fine-tuned models, model weights and checkpoints. Whether trained weights are protected by copyright is uncertain, so ownership and confidentiality should both be addressed expressly.
- Data: training, validation and test datasets assembled, cleaned or labelled by the CTO, together with any annotations and data lineage records.
- Prompts and configuration: system prompts, prompt libraries, hyperparameters and configuration settings, which may have significant commercial value.
- Documentation and know-how: technical documentation, model cards and records of how the model was trained, which are also increasingly needed for regulatory compliance.
Two further issues are common. First, an external CTO may bring their own pre-existing code, libraries or tools (background IP) to the project. The company will need either an assignment or a sufficiently broad licence of anything incorporated into its product. Second, the CTO may build the product using open-source software, open-weight models or third-party AI services. These are not owned by the company in any event, and their licence terms may impose obligations or restrictions that affect how the product can be used and commercialised. Where the CTO uses AI coding assistants, the question of whether AI-generated code is protected at all also arises, as we discuss in our separate article on ownership of AI-generated content.
What if the CTO works through their own company?
External CTOs often contract through a personal services company. If the CTO is employed by that company, the copyright in their work may vest in the service company rather than in the individual. The safest approach is for both the service company and the individual to be party to the agreement and to give the assignment, so that the rights pass to the client company whichever of them holds them.
Moral rights
Moral rights, including the right to be identified as author and to object to derogatory treatment of a work, cannot be assigned under the CDPA. They can, however, be waived by an instrument in writing signed by the person giving up the right. Under the CDPA, neither the right to be identified nor the right to object to derogatory treatment applies to computer programs, but a waiver is still advisable for other materials, such as documentation, designs and interface content.
How does the position differ in the US and the EU?
United States
In the US, the employer owns a “work made for hire” prepared by an employee within the scope of employment. Whether someone is an employee is determined by general agency principles, following Community for Creative Non-Violence v Reid (1989). Commissioned works created by a non-employee can be works made for hire only if they fall within nine categories listed in the Copyright Act and the parties agree this in a signed written instrument. Software is not one of those categories, so a contract that simply labels an external CTO’s code as a work made for hire will usually not be effective. A written assignment signed by the CTO is required, and US agreements typically combine a work-made-for-hire clause with an assignment as a fallback.
The wording of the assignment also matters. In Stanford v Roche (2011), the US Supreme Court left in place a ruling that an inventor who had “agreed to assign” inventions to a university had only promised a future assignment, and a later agreement under which he “does hereby assign” his rights to a company took priority. US agreements should therefore use present-tense language that assigns rights as they are created.
European Union
Under Article 2(3) of the Software Directive (2009/24/EC), the employer is entitled to exercise all economic rights in a computer program created by an employee in the execution of their duties or following the employer’s instructions, unless otherwise provided by contract. The rule does not extend to contractors. The Belgian Court of Cassation, for example, has refused to treat copyright in software developed by a company manager as having passed to the company where the manager had no employment contract and no written transfer had been agreed.
National laws also differ on how rights can be transferred. In Germany, copyright itself cannot be assigned, so the company must instead take a grant of exclusive exploitation rights. Some member states require contracts to specify each field of exploitation covered, so that a general statement that all rights belong to the company may not be sufficient. Where an external CTO is based in an EU member state, local advice on the form of the agreement may be needed.
What if the CTO is paid in equity?
External CTOs are frequently offered shares or options in place of, or in addition to, fees. Issuing equity does not transfer intellectual property. The IP assignment needs to be documented separately, and it is often sensible to make the signing of a full assignment a condition of the equity being issued or vesting. Leaver provisions should also be considered, so that the company retains the technology if the CTO leaves early. Tax-advantaged option schemes may not be available to non-employees, which may affect how the equity is structured.
How can gaps be fixed after development has started?
If development has started without proper documentation, the usual solution is a confirmatory assignment, in which the CTO (and any service company) assigns all existing and future rights in the work to the company, waives moral rights, confirms what background IP has been incorporated and agrees to sign any further documents needed. It is advisable to provide some consideration for the assignment, and to identify the materials covered as clearly as possible.
This is much easier to agree while the relationship is good. Once a CTO has left, or knows that a funding round depends on their signature, they may ask for payment or additional equity. Where the CTO will not cooperate, the company may need to rely on arguments about implied terms or equitable ownership, which are uncertain and fact-specific.
Securing ownership: a practical checklist
For businesses developing an AI product with an external CTO, securing ownership is likely to involve:
- Signed agreement from day one. Put in place a written consultancy or services agreement before work starts, containing a present assignment of all existing and future intellectual property in the work.
- Cover every AI component. Define the work broadly enough to include code, models, weights, datasets, prompts, configuration and documentation.
- Background IP. Require the CTO to identify pre-existing materials and grant a perpetual, irrevocable licence of anything incorporated into the product.
- Third-party components. Require disclosure and approval of open-source software, open-weight models and AI services used, and keep a record of the licences that apply.
- Include confidentiality obligations covering model weights, training methods, data and other know-how, and secure access to code repositories and cloud accounts in the company’s name.
- Service companies. Where the CTO contracts through a company, ensure both the company and the individual sign the assignment.
- Equity. Link the issue or vesting of shares or options to an executed assignment, and include appropriate leaver provisions.
- Close gaps early. Obtain confirmatory assignments for any past work well before a funding round or sale.
At Maybrook Law, we advise founders, investors and businesses developing AI technologies on the legal and commercial issues arising throughout the development and commercialisation of their products, including intellectual property, consultancy and technology development arrangements, and the use of third-party models, software and data. If you are developing an AI product with an external CTO or other consultants, or would like to discuss any of the issues raised in this article, we would be pleased to advise on your position.
This article reflects the law as at 29 August 2026. Our insights, articles and guides do not, and are not intended to, constitute legal advice or be an exhaustive review of all legal developments. Although every effort is made to ensure that the information provided in this article is accurate as of the publication date, please be aware that this area of law is developing rapidly and may be subject to change. Please seek legal advice before applying the information provided to any specific circumstances, transactions or legal issues.
