TSMC Pioneers a New Era in AI-Powered Trade Secret Management, Achieving Intelligent Innovation

TSMC is actively committed to intellectual property protection, and is continuously strengthening its core competitiveness through the Trade Secret Sustainable Intelligent Management Center. To further drive technological innovation and sustainable development, TSMC launched the new "Trade Secret Creativity Integration Partner (CIP) Intelligent System" in 2026. This system integrates Generative AI technology to conduct in-depth analysis of trade secret registration cases, assisting employees in uncovering innovative potential and simultaneously enhancing the technological value of "Golden Trade Secrets." This initiative sets a new paradigm for trade secret management, transitioning from "Preserving Innovation " to "Pursuing Innovation." TSMC also shared this achievement at the inaugural "Trade Secret Intelligent Management Expo and Workshop" in April of this year, attended by 22 companies that have established trade secret registration systems and over a hundred partners from other technology companies, promoting the industry’s prosperity together.

From Evaluation to Recommendation: Unlocking a New Mode of Human-AI Collaboration

Since establishing its trade secret registration system in 2013, TSMC has accumulated 670,000 registered cases, among which 3,858 have been awarded the prestigious "Golden Trade Secret Award." To further cultivate talent and spark innovation, TSMC developed the "Trade Secret Registration Innovation Index Analytics System" in 2023. This system was the first to use an AI model to identify inventors with innovative potential who had not yet received awards, and paired them with award-winning colleagues for one-on-one mentoring. By 2025, this initiative had nurtured 120 inventors with innovation-potential, 22 of whom successfully received the Golden Trade Secret Award.

Building on accumulated innovative achievements and talent cultivation experience, TSMC has once again made another breakthrough in 2026 by developing the "Trade Secret Creativity Integration Partner (CIP) Intelligent System," based on Generative AI technology. This system incorporates an in-house Large Language Model (LLM) and integrates Retrieval Augmented Generation (RAG) tools. It also utilizes Golden Trade Secret Award registration data for supervised learning training, ensuring the model possesses the ability to identify and create technological innovations that meet TSMC's high standards. It can precisely extract key technical features from different registration cases and analyze variations in their innovation levels. Through human-AI collaboration, the system effectively evaluates and ranks registration cases, generates specific explanations and forward-looking recommendations, and even offers perspectives previously unconsidered by the registrants. This inspires deeper, more comprehensive thinking while providing the evaluation committee with a recommended list of Golden Trade Secret Award candidates. The result is a significant enhancement in TSMC's competitive advantage in trade secret management and innovation.

By leveraging the "Trade Secret Creativity Integration Partner (CIP) Intelligent System," TSMC transforms static registration data into a dynamic source of innovation. Moving forward, TSMC will continue to focus on artificial intelligence (AI) and intelligent automation (IA) as foundational pillars, further improving the efficiency of trade secret management. This will enable TSMC to progress from generating innovative content to achieving comprehensive intelligent management, accelerating the momentum of the company’s sustainable development.

• What is Golden Trade Secret Award?

To recognize employees' significant impact on its competitive advantages (technology leadership, manufacturing excellence, and customer trust), TSMC annually bestows the Golden Trade Secret Award.

• What is Generative AI?

Generative AI creates novel content like text, images, music, or code, unlike traditional AI that primarily focuses on recognition or analysis. By learning from vast datasets, it masters generation rules and styles, enabling it to produce entirely new material based on prompts.