Can a competitor use the AI built with our company's information?

Suppose a manufacturer engages an outside developer to build a customized AI system to automate its quotation and cost analysis work.

The company provides the developer with product cost data accumulated over many years, pricing information for each customer, its quotation methods, internal operating manuals and other materials. The developer uses these materials to further train an existing AI model, or builds a system that searches the company's internal materials to generate answers.

What if, after development is complete, the developer seeks to sell the same system to other manufacturers, and even to the commissioning company's competitors?

The commissioning company may feel that it should have exclusive use of the system, since it not only paid substantial development fees but also provided its core business information and know-how. It would be hard to accept a situation in which competitors benefit from the know-how it built up without any comparable effort.

The developer's position is not simple either. The developer built the system using AI technology it already owned and its own development capabilities, and it cannot develop new technology from scratch for every client. If performing one company's engagement meant it could not use its existing technology, or the general technical experience gained during development, for other clients, its business could be significantly constrained.

Ultimately, the question is how far the developer may use the technology it legitimately owns while the commissioning company's information and know-how are protected.

Disputes over the copyright in, or the right to use, deliverables also existed in conventional outsourced software development. When AI is trained on a company's internal materials, however, there is an added issue. Even if the original materials are never copied as is or disclosed externally, the results of training on those materials can be used in systems built for other companies.

As AI-assisted outsourced work expands in the United States, a practice has emerged of separately addressing in contracts the permitted use of client data and the rights and liabilities relating to AI-generated output.

In reviewing a range of contracts between U.S. companies, including advertising and marketing services agreements, the author has also dealt with these AI-related issues firsthand. In actual contracts, the parties specify not only whether generative AI may be used, but also whether the client's confidential materials may be entered into external AI tools or used to train models, whether prior approval is required to include AI-generated material in deliverables, and who bears liability for that material.

It is especially noteworthy that these contracts distinguish between entering client materials into an AI tool and using them to train an AI model. Using a client's materials for a specific task is a different matter from using those materials to improve an AI's performance and then using them for other work in the future.

Actual disputes are also arising. In 2025, a lawsuit was filed in the United States over allegations that the design software company Figma had used its users' designs and data to train its AI products without authorization. The case does not involve a dispute over ownership of customized AI deliverables, but it is noteworthy in that whether a vendor may use materials provided by its service users for its own AI development became the subject of a dispute. The court, however, only compelled arbitration and stayed the lawsuit, and did not rule on whether the use of the data itself was unlawful.

These U.S. contract practices are an important reference when reviewing outsourcing agreements of Korean companies as well. Rather than adopting U.S. contract provisions as they are, however, the parties need to define their rights and liabilities specifically, taking into account Korean law, the actual development method and their business relationship.

How, then, should a dispute in Korea between the manufacturer and the developer discussed above be decided?

First, a commissioning company does not automatically acquire exclusive rights to the entire AI system merely because it paid development fees and provided internal materials. The technology the developer already owned, the newly developed programs and the results of training on the company's materials need to be examined separately as a legal matter.

Under Korea's Copyright Act, computer programs can be protected by copyright, but the copyright in a program does not automatically vest in the commissioning company merely because it paid outsourced development fees. The specific rights must be determined by examining the actual development process, the terms of the contract and other factors. Nor are the results of training an AI model always recognized as an independent copyrighted work. In the end, what matters is what was developed and which rights the parties agreed to transfer or license.

Conversely, the fact that the developer holds rights in the AI system does not mean it may freely use the commissioning company's business information.

If information provided by the company, such as cost data, pricing policies or manufacturing processes, qualifies as a trade secret under the law, it can be protected under the Unfair Competition Prevention and Trade Secret Protection Act. Even if it does not meet the requirements of a trade secret, if the contract designated it as confidential information or prohibited its use for anything other than a specific purpose, that agreement becomes an important standard for judgment.

The mere fact that the developer used the company's materials to train an AI does not mean that every subsequent act of providing that model to other clients immediately constitutes trade secret misappropriation. However, even without directly disclosing or copying a trade secret, using it to save the time and cost required for development, for example, may constitute use of the trade secret. Ultimately, the protected information, how it is actually used and the scope of use permitted under the contract must be examined in detail.

Here, the characteristics of AI technology make the issue even more complex. If the AI model itself was further trained on the company's materials, the results of that training may remain in the model even after the original materials are deleted. On the other hand, if the system stores the company's materials in a separate database and searches them when needed, the model itself may not have been trained on those materials. The protection of the information and how it is handled after the contract ends will inevitably differ between these two cases.

What matters, therefore, is not simply deciding who owns the AI. It is distinguishing how the company's information is used by the AI and how far the developer may use the results for other clients.

When reviewing contracts of Korean companies, we often find that these issues are not yet sufficiently addressed. The contracts contain conventional copyright ownership and confidentiality provisions, but do not clearly address the use of client materials for AI training or the reuse of the results for other clients. Because such gaps in the contract can lead to future disputes, it is becoming increasingly important in actual contract review to define the related rights and liabilities specifically according to how AI is used.

For example, vesting the copyright in the development deliverables in the commissioning company does not fully resolve the question of whether the developer may use the AI models or technology it already owned for other clients. Conversely, the parties could consider letting the developer keep the rights in the model while contractually restricting it from providing to competitors the results of further training on the commissioning company's materials.

The same is true of confidentiality of client materials. Prohibiting external disclosure must be distinguished from prohibiting use for AI training. Also, simply providing for the return or deletion of the original materials may not fully resolve how information already used in training is to be handled.

That said, prohibiting all reuse by the developer is not the answer. Restricting even the developer's existing technology and general development know-how could raise development costs or make it difficult to reach an agreement at all. A more realistic approach is to distinguish the information the commissioning company actually needs to protect, and the deliverables it needs to use exclusively, and to define the scope of reuse accordingly.

In particular, once development is complete, it can become difficult to tell which technology belonged to the developer from the outset and which parts were created using the commissioning company's materials. These rights and the scope of use therefore need to be settled at the contracting stage, not after the deliverables are completed.

In this sense, an AI outsourced development agreement is becoming more than a contract for receiving development deliverables. It is becoming a contract that determines who may use the information a company provides, and the results derived from that information, and for what purposes.

As AI is put to use in companies' operations, what outsourced development produces is also changing. In the past, the main concern was who owned the programs or deliverables completed by the developer. Now, how the data and know-how a company provides are used by AI, and how far the results may be reused, have also become important questions.

The information a company has built up over many years is, in itself, an important competitive asset. As AI technology expands the ways such information can be used, its protection also needs to be approached differently than before.

What matters is not restricting the use of AI technology itself, but distinguishing the information the company provides from the technology the developer owns, and defining the scope of their use reasonably. Whether other companies can use an AI that a company developed with its own materials may ultimately depend greatly on the rights and liabilities the parties agreed on before development began.