Enveil Secure AI is a Privacy Enhancing Technology (PET)-powered solution for securing AI and machine learning workflows. It is built on Enveil's ZeroReveal Machine Learning platform and addresses risks associated with training, evaluating, and validating ML models on sensitive or proprietary data. The solution operates by enabling encrypted ML operations, allowing models to be trained, validated, and evaluated without exposing the underlying data or the model itself. This approach allows organizations to work across organizational, jurisdictional, and third-party boundaries without compromising data ownership or security. Key operational capabilities include: - Encrypted evaluation and inference: ML models can be used for inferencing without revealing model contents or inputs. - Encrypted model training via federated learning: Models can be trained across decentralized, disparate datasets while keeping data in place and encrypted. - Encrypted model validation: Models can be validated without exposing sensitive data used in the validation process. The solution targets the risk that ML models encode the data they are trained on, making them vulnerable to adversarial attacks that can extract sensitive information. By applying PETs, Enveil Secure AI aims to mitigate these adversarial ML risks while still allowing organizations to derive actionable insights from broad, diverse data sources. Use cases span commercial and government/federal sectors, including national security and multi-domain operations, as evidenced by a U.S. Army contract.