Model theft is a type of cyberattack where an attacker illicitly copies or replicates the underlying structure, data, or behavior of an artificial intelligence (AI) model. By exploiting vulnerabilities in how AI models are accessed or interacted with, attackers can reconstruct the model’s logic, parameters, or training data. This enables them to use, modify, or resell the stolen model, potentially bypassing intellectual property protections and security controls. Model theft can lead to loss of competitive advantage, exposure of sensitive data, and increased risk of further attacks.