INTRODUCTION Contributed by: Simon Bushell, Gareth Keillor and Maitreyee Dixit, Seladore Legal
Global Overview As businesses around the world evaluate their options for protecting valuable intellectual prop - erty in the context of today’s dynamic techno - logical environment and highly mobile labour force, trade secret protection can be an essential complement to patent, copyright and trade mark protections. This is particularly true in the USA in light of recent developments in the patent system – including shifting judicial standards for patent- eligible subject matter and the increased avail - ability of post-grant challenges at the patent office – that have increased the importance of trade secret protection as an alternative vehicle for protecting intellectual property. Moreover, as the developed world continues its shift from a manufacturing economy to a knowl - edge-based one, where the most rapidly grow - ing sectors offer software and services, trade secret laws are more relevant than ever. Artificial Intelligence in Full Force Generative artificial intelligence (AI) is here to stay. Various industries have begun using large language models (LLMs) to analyse big data, create work products and even innovate by developing novel ideas or inventions. AI applications and LLMs raise several issues for trade secret protection. First, they may cap - ture and store information that may be used to train and enhance the AI’s ability to gener - ate results. If one were to input a trade secret into an AI application or LLM prompt, the trade secret could be at risk of unintended exposure to the company behind the AI application depend - ing on the terms of the application’s end-user licence agreement. This concern is particularly salient in light of the expanded use of generative
AI in the workplace, which has resulted in dis - closures of trade secrets through ChatGPT and Sundstrom’s leak of confidential meeting notes and data through the AI tool “Otter”. Further - more, creators of some of the largest generative AI applications, such as OpenAI, preserve the ability to review inputs provided by users and potentially disclose such inputs to affiliates or third parties. Second, the trade secret could be used as a training input for other problems or prompts, resulting in potential exposure to oth - er end users of the AI application. Third, trade secrets stored by the AI application, which often occurs based on AI applications’ storage of training inputs provided by their users, may be at risk of exposure from security breaches tar - geting the companies behind the AI application. Each of these issues will push trade secret own - ers to implement new ways to safeguard their trade secrets, such as updating employment agreements, drafting internal AI-use policies that limit the ways in which employees may use generative AI, training employees in light of such updated policies and agreements and carefully negotiating with companies behind AI applica - tions to limit the use or accessibility of trade secret inputs. Alternate arrangements to enable greater trade secret protection may include the purchase or development of an internal genera - tive AI application or use of custom non-disclo - sure agreements for generative AI tools. Another evolving interaction between generative AI and trade secret protection concerns scenar - ios in which generative AI itself produces trade secrets. Unlike patent and copyright protection, trade secret protection is not limited to human inventors, and the broad definition of “trade secret” may enable protection of entire AI plat - forms, training algorithms, inputs and outputs. The scope of protection for AI-generated trade
6
CHAMBERS.COM
Powered by FlippingBook