Shadow AI: The Murky Threat to Enterprise Adoption of Generative AI

This article analyzes the rise of Shadow AI in enterprises and associated risks—especially data leakage and compliance violations—citing examples like Samsung’s ChatGPT data disclosures and Microsoft’s GitHub exposure; it argues traditional sanitization and anonymization are insufficient, surveys privacy-enhancing technologies (homomorphic encryption, MPC, differential privacy), and promotes confidential computing with trusted execution environments as the basis for a “confidential AI” approach that protects data in use, supports model training and inference on sensitive data, and strengthens regulatory compliance.