Businesses using artificial intelligence (AI) to reshape industries want to protect their AI models. Patent registration is one option for safeguarding intellectual property (IP), but may not always be feasible for AI models. The post-Alice landscape rendered many software-related inventions abstract ideas, making them unpatentable. See Alice Corp. v. CLS Bank Int’l, 573 U.S. 208 (2014). And some AI models use off-the-shelf algorithms that may not meet the novelty requirement for patentability. Further, if IP is not primarily created by a human, it may not qualify for patent protection. See Inventorship Guidance for AI-Assisted Inventions, 89 Fed. Reg. 10234 (Feb. 13, 2024). 

What Other Protection Is Available? 

Trade secrets are a viable option for protecting AI models. Trade secrets offer a versatile alternative for protecting a company’s core algorithms, data, processes, techniques, and more against its competitors – privately, quietly, and indefinitely. By leveraging trade secret strategies, businesses may avoid the need for the public disclosure required by patent filings. 

What Is A Trade Secret?

A trade secret is any information, such as an algorithm, code, device, design, formula, method, process, procedure, or technique that (1) the owner has taken reasonable measures to keep secret, and (2) holds independent economic value, actual or potential, from not being generally known or readily ascertainable by others who can obtain economic value from its disclosure or use. 18 U.S.C. § 1839(3). Unlike other forms of IP, trade secrets are not formally registered; instead, they are protectable so long as secrecy is maintained. Failing to take reasonable measures to protect the information as secret may result in the loss of its status as a trade secret. After all, something that is public cannot be a secret. 

Ways to Protect AI-Generated Intellectual Property

While AI models may come in different shapes and sizes, the following measures can help safeguard a model as a trade secret: 

Define Protectable Aspects of the AI Model

The first step in protecting an AI model as a trade secret is to define the protectable aspects of the model. There are different aspects of a training model, which may be subject to protection, e.g., underlying data, the AI algorithm, the finetuning process, etc. It is important for the owner of the model to clearly decide what aspects of the model are protectable and make conscious efforts to protect these aspects. For example, if the underlying data is protectable, it must be documented and access to the data must be restricted. 

Restrictions on Using the Model

Most AI models are delivered as software-as-a-service (SaaS) solutions, with licensing agreements that specify the permitted uses and intended purposes of the models. It is important to limit the scope of these licenses so that the model is not vulnerable to attacks or reverse engineering. 

Security Protocols

Secure storage of trade secret materials is essential, whether they are in digital or physical form. Digital materials should be protected with strong passwords, encryptions, and muti-factor authentication tools, while physical materials may be stored in safes, locked cabinets, or other secured areas away from the general public and employees without clearance. Further, access controls should be implemented to prevent unauthorized viewing, sharing, or altering of the information in question. Technical measures, such as model watermarking, fingerprinting, and obfuscating should be implemented to trace any unauthorized access and make reverse engineering more difficult. 

Authorized Personnel 

To maintain the confidentiality of the IP in question, access to the information should be limited to authorized personnel only. These authorized personnel should be individuals whose roles specifically require access to the information to perform their job duties. Access by others, e.g., investors, should rarely be granted and only if the accessing party signs a non-disclosure agreement. 

Employment Agreements

Authorized personnel with access to trade secrets should sign employment agreements that include clauses outlining their obligations to maintain secrecy and specifically require them to disclose at the beginning of their employment any confidential or proprietary materials they intend to use in their new employment. Further, employers may take actions to further ensure compliance by sending out regular reminders and renewals of appropriate agreements to ensure ongoing compliance and awareness in safeguarding the applicable trade secrets. 

Monitoring Tools and Audits

Despite best efforts, it is always possible that confidential aspects of an AI model may be accessed by unauthorized users. It is therefore crucial to deploy use-monitoring tools to track data access and flag suspicious activities. It is also recommended to conduct regular audits to ensure that both employees and security measures are prepared to protect confidential aspects of the AI model. 

Confidentiality

Whether the materials are in digital or physical form, it is important to designate them as confidential wherever possible. The underlying information, source codes, training data, parameters, and other proprietary information should not be publicly disclosed. Disclaimers, such as “CONFIDENTIAL FOR AUTHORIZED PERSONNEL ONLY,” may be implemented on physical materials to further inform employees of the sensitive nature of the information contained. 

Action Plan

In the event of an actual or suspected trade secret breach, the company should develop a comprehensive action plan detailing the necessary steps to effectively address the situation. This plan should include guidance on identifying and tracing the breach, containing the damage, contact information for key personnel, and available actions at various stages to protect the company’s trade secret information and mitigate any potential harm. 

Training and Trade Secret Policy

Employees may not be familiar with best practices in safeguarding IP. Therefore, it is critical that an employer offer an easy to understand trade secret policy and conduct regular training on handling confidential information. It is best to write the policy with as little legalese as possible to ensure employees understand it. The training should cover trade secrets generally to create a strong foundation for each employee’s understanding. Further, training sessions should cover the importance of maintaining confidentiality, potential consequences of breach, and specific protocols in place to protect the trade secrets. Finally, the trade secret policy must be easily accessible.

Departure Protocol

Employees’ duties in handling trade secret information should not end with their employment. Instead, clear departure protocols need to be established to ensure that departing employees return or destroy any trade secret materials in their possession and are reminded of their ongoing obligations to maintain confidentiality of the information.

Monitoring

As AI is an evolving area, staying informed is crucial. Beyond the recommendations set out above, businesses should actively monitor AI-related litigation and update their internal policies to ensure robust protection.

---

Originally published on August 1, 2025 online with Federal Bar Council Quarterly. Reprinted with permission. Further duplication without permission is prohibited. All rights reserved.