Protecting Trade Secrets: Tips for AI Companies


3 minute read | January.22.2025

AI companies should familiarize themselves with trade secret law to safeguard their innovations.

A company does not need to register a trade secret to invoke it in litigation, unlike other IP protections. Trade secrets can protect algorithms, processes, datasets and more.

Their ease of creation and flexibility make trade secrets especially important to protect for companies that focus on AI, where the tech changes quickly and other IP protections may not apply.

The Basics: Trade Secrets 101

Federal law says a trade secret must relate to secret information that “derives economic value, actual or potential, from not being generally known.” In addition, the owner must have “taken reasonable measures to keep such information secret.”

Trade secrets do not protect IP absolutely. Rather, they protect trade-secret-eligible IP from being misappropriated. This means that if a company discovers a rival’s trade secret through “fair and honest means” it loses protection.

Here’s a look at how AI companies can protect trade secrets and avoid infringing upon those of others.

A Closer Look: The Three Elements of a ‘Trade Secret’

The law outlines three requirements for a “trade secret.”

Secret information: The sheer breadth of information that may constitute a trade secret makes it a compelling tool for protecting IP.

  • Courts have found some AI-based algorithms, training data and data-harvesting processes eligible for trade secret protections.
  • Even effective prompts for large language models like ChatGPT could potentially be trade secrets.

Not generally known: To qualify as a trade secret, the “actual or potential” economic value of the information must stem from its secrecy. This is a simple standard to meet for breakthrough technologies like AI.

Reasonable efforts to maintain secrecy: Companies can automatically meet the first two criteria for a trade secret, but maintaining secrecy requires action.

  • A classic example of maintaining secrecy comes from The Coca-Cola Co.’s efforts to protect the formula for Coke.
  • The company limited to two the number of people who knew the formula. It also locked the only copy of the formula in a vault that required a resolution from the board of directors to open.

How AI Companies Can Establish Secrecy

Label information as confidential.

  • Whether you want to protect your model architecture, training data curation processes, algorithms, training and validation data or even the prompts employees use, make sure to clearly label information confidential.
  • Companies can do this by marking the information itself or through more general means, like requiring employees to read internal policy guidance.

Limit internal access.

  • Consider limiting access to employees whose work requires it. Implementing password protection and other cybersecurity protocols can go a long way toward establishing secrecy.

Limit external access.

  • Requiring third parties to sign non-disclosure agreements, data or model access agreements and using restrictive licenses can show a company has taken reasonable steps to protect a trade secret.
  • This can come up for AI companies when external researchers work on aspects of R&D or product validation. For example, many producers of language models allow researchers some access to their products to test for harmful or biased outputs. Overly broad access risks a company losing trade secret protection.
  • Companies should take care when publishing findings relevant to their product, including at conferences.

Establish secrecy on the way in.

  • Consider including an onboarding agreement that assigns employee-generated IP to the company. That can establish general protection around company IP.
  • Often called confidential information and invention assignment agreements, they typically contain provisions around non-disclosure, non-competition and non-solicitation of company customers. They also may include an agreement that employees will return company property when their employment ends.

Quickly and thoroughly discontinue access to a departing employee.

  • Ensure an employee who leaves the company returns all materials that may include the potential trade secret. Otherwise, delete the information.
  • If an employee leaves for a competitor or may pose a trade secret risk, consider conducting a forensic search to verify the employees doesn’t take company information.
  • Exit interviews are a good way to uncover whether heightened diligence is necessary.

How to Protect Yourself From a Trade Secret Claim

Trade secret claims are bound to arise as employees migrate between companies working on cutting edge products. Deploying a product that incorporates another entity’s trade secret is not inherently a violation. Independent invention, accidental disclosure or reverse engineering each provide a complete defense to a claim of trade secret appropriation. 

Consider this example: Company A uses publicly available research from Company B. It also has authorized access to Company B’s product. Company A discovers IP that is otherwise protected by Company B. In that case, Company A has not misappropriated Company B’s trade secret. This is true even if Company A deploys that technology. Independent discovery is a complete defense to a trade secret appropriation claim.

On the other hand, if an employee from Company B divulges trade secrets during a job interview with Company A, and Company A hires that employee, Company B will have a stronger claim.

While this may seem obvious, many companies require presentations of past work as part of their interview process – a practice that can easily replicate our example scenario.

To mitigate potential liability, consider implementing real-time evaluations, such as technical interviews, instead of examining a candidate’s past work. 

Since the manner in which a trade secret is discovered determines whether a claim is viable, not the technology itself, trade secret actions tend to be highly fact-dependent and can come up often in industries that depend on emerging technology.

Summing Up

As AI evolves and become integral to various industries, understanding and leveraging trade secret law is an essential way for companies to protect their innovations.

Companies can safeguard their competitive edge by maintaining the secrecy of valuable information, such as by implementing robust internal policies, limiting access and securing agreements with third parties.

To avoid legal pitfalls, it’s also crucial to keep in mind how people or organizations can inadvertently disclose or misappropriate trade secrets.

Want to know more? Contact one of the authors (Shannon Yavorsky, Robert Shwarts or Tom Zick).