A security threat where an attacker attempts to steal or duplicate an AI model's functionality by repeatedly querying it and analyzing its responses.