<p>Advances in image synthesis now enable the creation of hyperrealistic portraits of real individuals, including celebrities and private persons. This shift raises sociotechnical concerns that go beyond perceptual responses to virtual characters or robots, particularly regarding identity, consent, and reputational vulnerability. This article introduces the concept of AI-mediated identity proximity, the extent to which a synthetic image is tied to a real person, and argues that identity-bearing synthetic images require distinct theoretical and governance perspectives. To offer empirical grounding, we conducted a small exploratory pre-test (<i>N</i> = 26) comparing synthesized celebrity faces and self-images across morph levels. Celebrity images were generally perceived as more likable and less discomforting, yet perceived reputational risk increased similarly for both categories, suggesting that individuals recognize the vulnerability of any real identity in synthetic media environments. Building on these preliminary insights, we propose a sociotechnical framework explaining how AI-mediated identity proximity influences affective responses, agency perceptions, and vulnerability judgments. We conclude with design and policy implications for digital consent, synthetic identity governance, and responsible deployment of facial synthesis technologies.</p>