We propose a new approach in data security, known as HbHAI (Hash-based Homomorphic Artificial Intelligence). This disruptive approach enables processing data under their encrypted form without the inherent limitations of conventional homomorphic data analysis techniques to date (CKKS, BFV and TFHE schemes). HbHAI is based on a new class of key-dependent hash functions that naturally preserve the similarity properties, most AI algorithms rely on. HbHAI techniques are not yet available to the public, as the question of industrial valorization and protection is not yet fully resolved. However, to enable first public feedbacks, this document presents the first formalization of these techniques. Among their many features, HbHAI techniques can reduce data size with a compression ratio of at least 3. While strongly preserving data security and confidentiality, this reduces storage space and computation time for native, “ready-to-use” AI algorithms.