It is no secret that large language models (LLMs), such as Open AI’s CHATGPT or Google’s Gemini, have become commonplace in our contemporary business world. Many forward-thinking companies have delegated to AI to save time, create operational efficiency, and to allow human workers to focus on more strategic and creative endeavors, with the hope that this will propel their business forward. These tools can make life easier by automating tasks such as taking notes or summarizing meetings, drafting emails or other communication, performing deep research, processing large volumes of data, running complex calculations, etc. For teams focused on growing their businesses, using these tools seems like a no-brainer, but many don’t realize the risks associated with providing private or sensitive company material into open LLMs.
When information is shared with AI platforms, there’s a chance it can be accessed, stored, or even used by the platform in responding to other users’ requests. Eventually being presented to another user in answering their submitted prompt. Naturally, this can lead to potential legal problems, especially when a company shares confidential information, such as trade secrets, and depends on keeping such information secret and protected. If you’re not thinking ahead, this information can be unknowingly shared outside the company, thus potentially negating the sole purpose of having a trade secret.
Open-source vs Proprietary AI Models, and What They Mean
Understanding what type of AI model you or your business uses is fundamental to understanding the associated legal risks. A “model” can be thought of as the brain of artificial intelligence. These virtual brains are fed data to learn and carry out the tasks they are designed to do. Although these models can be built to complete any kind of task, most fall within two broad categories: open-sourced and proprietary.
These two frameworks fundamentally differ on three points: licensing, control, and customization. Open-sourced models are characterized by the ability for users to have free access, unrestrained distribution rights, and the freedom to edit publicly available source code. On the other hand, proprietary models allow access usually through paid subscriptions, operate in a closed-circuit environment, and bar the user from editing any source code—instead relying on pre-made functions. In short, open-sourced models champion flexibility and offer up the ability for businesses to create tailor made solutions themselves. Proprietary models are much more established, offering scalability and pre-baked resources.
All AI models have their own specific method of growth—meaning how the model relays answers, collects information, and digests user submitted prompts. Thus, diligent understanding of the tools your business is using is paramount to maintaining adherence to the definition of a trade secret.
AI and Trade Secrets
It’s also worth having a discussion about trade secrets, which are a subsection of the field of intellectual property (IP) law. For information to qualify as a trade secret, it must: (1) derive independent economic value from not being generally known or readily obvious; (2) it must be subject to reasonable efforts to maintain its secrecy; and (3) it must provide an advantage to the business over competitors who do not know or use it. Trade secrets don’t require registration, instead relying on confidentiality agreements or individuals limiting access to the information. Trade secrets can range from financial information to manufacturing or other business processes to customer information, and so on.
When considering both the framework of various AI models your business uses, and the trade secrets you may have, caution should always be taken. A thorough review of how your AI tools collect, store, and use your data will give clarity on what to include and not include when interacting with the AI. If your AI model is built to learn and “grow” from what users give them, that means when someone enters meeting notes, code, data, or internal strategy into a program, it might not stay private. These platforms use the data to get “smarter,” and sometimes that means the information submitted ends up being saved or applied in ways you can’t control.
Even using AI programs to record meetings and take notes comes with similar risks. You may be exposing internal communications to external parties, meaning you need to manage how and when you use AI. It’s probably best to retain internal tracking when necessary and apply those practices to any conversations involving protected information.
All Employees Should Agree to Keep Trade Secrets Out of AI Submissions
It’s one thing for leadership to understand the risks. But unless every employee is on the same page, the risks will not go away. Employee agreements and policies should clearly state that confidential business material should not be disclosed without authorization, and that includes entering such information into AI tools unless there is express approval from management.
These details shouldn’t be buried in fine print. It should be discussed in employee training workshops, dictated in employee handbooks, and sufficiently addressed during the employee onboarding process. Everyone should know what is deemed as confidential, and what the limits are to using that data. People don’t always mean to overshare, but they won’t know where the line is if you don’t discuss the expectations.
Trade secret laws only work if a company is able to demonstrate that it made appropriate efforts to keep the information private. If a worker uploads a private client list or business strategy into ChatGPT to get help writing an email, that could be enough to fail at the ability to make such a showing.
Protecting Your Innovation Amid Advanced AI
AI tools are here to stay as a regular resource for innovative businesses. Using them wisely doesn’t mean avoiding them altogether. It means setting clear boundaries, updating policies, and ensuring best practices to protect the innovation of your business. If your company handles proprietary ideas, products, or methods, you need strong legal protections in place. Contact McDermott IP Law to put those protections in place and ensure your trade secrets survive as technology continues to advance.
McDermott IP Law
Latest posts by McDermott IP Law (see all)
- Can You Trademark a Sound, Smell, or Color? Here’s What to Know - January 15, 2026
