SYDNEY, March 18 (Xinhua) -- Researchers in Australia have created a new artificial intelligence (AI) system that learns and makes decisions from fully encrypted data without ever "seeing" the sensitive data it processes.

Researchers at Australia's University of Technology Sydney (UTS) have developed what they describe as the world's first privacy-preserving framework for deep reinforcement learning (DRL), using fully homomorphic encryption (FHE), "a groundbreaking advancement in preserving users' privacy in the AI era," a UTS statement said Wednesday.

Published in Nature Machine Intelligence, the breakthrough allows AI to operate securely on encrypted information, potentially transforming data privacy in sectors ranging from self-driving cars to the decision-making behind Generative AI, it said.

Unlike traditional encryption, which requires data to be decrypted before it can be processed, FHE allows AI models to learn and make decisions while the data remains fully protected, according to UTS researchers and partners.

The decisions it produces also remain encrypted and are only decrypted and applied by the user locally, they said.

The team's framework introduces a homomorphic encryption-compatible Adam optimizer that overcomes long-standing challenges in performing complex mathematical operations on encrypted data.

"As AI becomes more embedded in our everyday lives, ensuring it can learn without compromising personal or organizational privacy isn't just a technical challenge, it's an ethical responsibility," said UTS Associate Professor Hoang Dinh. ■