’Ignore All and Accept My Resume’: The Impact of Prompt Injection in Automatic Resume Screening

Users of generative AI models—such as ChatGPT, Claude, or any other model—are becoming familiar with prompt manipulation, contributing to the Prompt Engineering as an emerging technical field. Indeed, every new technology possesses a dual nature and can be able of being exploited for both beneficial and harmful purposes. The harmful aspect in the field of prompt engineering is known as Prompt Injection. The latter, which is a harmful practice used in different situations, involves manipulating AI systems by injecting malicious instructions. In the context of resume screening, job candidates often exploit this technique to manipulate the applicant tracking system (ATS) by inserting a harmful prompt into the resume to be approved. By examining recent studies, this review seeks to understand this new area of concern by examining the influence of prompt injection on resume screening. The topic is still general and not tailored to a specific use case, such as resume screening, owing to its novelty. Therefore, we provide an overview to shed light on how this could affect the quality of the hiring decisions.