The increasing presence of intelligent machines in workplaces raises important questions about how human workers respond to them. Mind perception theory suggests that machines with emotional capabilities (high experience) elicit aversion through the uncanny valley effect, while research on human-automation interaction predicts that highly competent machines (high agency) threaten humans by evoking replacement concerns. This research reconciles these seemingly contradictory perspectives by examining how different dimensions of machine minds, agency (thinking intelligence) and experience (feeling intelligence), influence human identity at work and subsequent unethical behaviors toward machines. Across two vignette experimental studies (N=272 healthcare employees; N=285 hospitality/retail employees), we uncover a paradox: contrary to mind perception theory's predictions, 'warm' machines (perceived with advanced feeling intelligence) enhance employees' identity at work, which in turn reduce unethical behaviors toward machines in the form of machine-directed sabotage, while 'smart' machines (perceived with high thinking intelligence) threatens employees' identity at work, which in turn increase machine-directed sabotage. By showing that different types of machine intelligence trigger distinct psychological processes (i.e., thinking intelligence threatening identity and feeling intelligence enhancing identity), this research advances our understanding of how distinct forms of machine intelligence elicit identity-based responses that drive unethical behavior toward machines. Our findings offer theoretical insights into human-machine interaction and practical guidance for organizations seeking to integrate intelligent technologies in ways that promote ethical and cooperative workplace behavior.