Companies face mounting uncertainty over whether proprietary information remains legally protected as a trade secret once it is input into artificial intelligence systems by employees.
The challenge stems from a fundamental mismatch: trade secret law requires companies take reasonable measures to maintain confidentiality, but that standard predates the widespread use of generative AI tools that employees now deploy daily for coding, research and business processes, according to IP attorneys and legal experts. The uncertainty is creating tensions between productivity gains and intellectual property safeguards that existing law may not adequately address.
Courtney Lytle Sarnow, a partner at CM Law who specializes in IP, noted trade secrets face unique vulnerabilities compared to patents and copyrights.
"Trade secrets are usually overlooked — they're kind of the afterthought of the IP world," she said. "But they may be some of the most valuable [protections] because they will protect equally the technology of humans who are creating these systems and the technology created by these systems."
The tension matters because trade secrets often protect companies' most valuable assets, including manufacturing processes, customer lists, algorithms and business strategies that provide competitive advantages but do not qualify for patent protection. Unlike patents and copyrights, trade secrets have no registration process or examination by regulators. Protection exists only as long as companies maintain secrecy through reasonable efforts.
Disclosure through AI
When employees enter proprietary information into large language models such as Anthropic PBC's Claude or OpenAI LLC's ChatGPT, coding assistants or other AI tools, legal experts say it is unclear whether companies have disclosed their trade secrets — even if the AI model does not reproduce the exact text.
"The question becomes how much human contribution would be required to maintain protection," said Steve Wang, a partner at Caldwell and an expert in IP litigation, adding companies need to "document everything" showing they conceptualized inventions rather than simply prompting AI systems.
Maryam Meseha, a partner at Pierson Ferdinand who advises companies on AI governance, said the core issue is twofold: whether AI-generated output can qualify as a trade secret under traditional definitions requiring human authorship, and whether inputting proprietary information into open-source models destroys confidentiality protections.
"If it's out in the open and it's in the ether, is it still considered a secret?" Meseha said. "Companies that are getting it right are treating AI as a governance piece — clearly documenting and assessing how models are trained, figuring out how they're using it internally, and then putting controls in place to have a human in the loop."
Reverse engineering at scale
Trade secret law traditionally allows competitors to reverse-engineer publicly available products. AI is changing the scale and speed of that process.
Wang said if someone uses AI tools to reconstruct a trade secret without accessing confidential information or misappropriating it, "they essentially reverse engineer it on their own", potentially creating a valid defense against infringement claims.
Sarnow said whether AI-powered reverse engineering defeats trade secret protection remains legally unsettled.
"If AI breaks open your trade secret, does that ruin the protection?" Sarnow said. "That's something that we won't know for a while."
Attribution challenges
When AI generates code or processes matching competitors' trade secrets, liability becomes murky. Legal experts say courts have no framework for determining responsibility when AI serves as an intermediary.
The United States Patent and Trademark Office attempted to address related issues in November 2025 by revising inventorship guidance to reiterate that "only humans can be the inventors or authors," Wang said. But determining "how much human contribution is needed when we're treating AI as a tool" still happens on a case-by-case basis, he noted.
Courts have consistently rejected non-human authorship in IP cases. A federal appeals court ruled in March 2025 that a machine cannot be listed as an author on copyright applications, with the court stating the Copyright Act "requires all eligible work to be authored in the first instance by a human being."
Trade secret law presents a similar challenge. While AI systems can generate economically valuable information, owners must still articulate and describe that information to pursue misappropriation claims, according to an analysis by the Brookings Institution.
The path forward
Companies are tightening documentation protocols, while recognizing that perfect protection may be impossible, experts said. A 2023 Burkhalter Kessler legal analysis suggested banning AI tools entirely is the only way to ensure trade secrets remain confidential, as Samsung Electronics Co. Ltd. did for its employees following reported leaks of sensitive data via ChatGPT.
Other firms such as Jones Day and Winston & Strawn contend that such a ban would not only be impractical but costly, and risk putting businesses at a competitive disadvantage. Instead, they suggest limiting access to sensitive data and purchasing or developing their own internal GenAI applications.
Meseha said her clients are focusing on internal governance by using closed AI systems rather than open-source models for sensitive work, establishing clear parameters for AI use, and maintaining human oversight to preserve claims of human authorship.
"This is a lifecycle," Meseha said. "These considerations have to continually be monitored. It's not a point in time."
Regulatory clarity remains distant. Meseha noted the Trump administration has prioritized AI innovation over regulatory frameworks, though national security and IP protection concerns are creating pressure for baseline safeguards.
"There has to be kind of a marrying of innovation and regulation," she said, "and I think we're still trying to figure that out."
What's happening this week?
Below is a list of hearings, webinars and other tech, media and telecom-related events taking place virtually and in person in the nation's capital and beyond this week:
Feb. 9
➤ House Committee on Rules: To meet on Undersea Cable Protection Act of 2025, Securing America's Critical Minerals Supply Act
Feb. 10
➤ S&P Global Market Intelligence: Datacenter Dealmaking: Can the Record-Breaking Momentum Continue in 2026?
➤ American Enterprise Institute: 30 Years of the Telecommunications Act — What Did We Learn?
➤ Senate Appropriations Subcommittee on Commerce, Justice, Science, and Related Agencies: Hearings to examine a review of broadband deployment funding at the United States Department of Commerce.
➤ U.S. Senate Committee on Commerce Science and Transportation: Hearings to examine media ownership in the digital age.
Feb. 11
➤ National Telecommunications and Information Administration: Listening Session on the Use of BEAD Funds
➤ House Education and the Workforce Subcommittee on Workforce Protections: Building an AI-Ready America — Safer Workplaces Through Smarter Technology
➤ House Science, Space, and Technology Subcommittee on Research and Technology: Accelerating Progress — US Surface Transportation Research
Feb. 12
➤ The Brookings Institution: A conversation with NTIA Assistant Secretary Arielle Roth on the future of the digital economy
➤ International Center for Law and Economics: Modernizing Video Competition — Policy Solutions for the Digital Age
Feb. 12-13
➤ World AI Cannes Festival
