Shadow AI: How to Mitigate the Hidden Risks of Generative AI at Work
By Digital Jewels
Published On July 30, 2025
Once proprietary data is processed by a public AI tool, it may become part of the model’s training data, serving other users down the line. For example, in March 2023, a multinational electronics manufacturer was reported to have experienced several incidents of employees entering confidential data, including product source code, into ChatGPT. Generative AI applications, such as large language models, are designed to learn from interactions. No company wants to train public AI apps with proprietary data. | Shadow AI
