Artificial intelligence (AI) is increasingly used in medicine, making it one of the largest and fastest-growing fields in which AI is applied. Previous research has shown that experts in the domain in which AI is used reported lower levels of trust in the AI’s suggestions. There is, to our knowledge, no research investigating whether the risk-benefit perceptions of experts in the domain in which AI is applied differs compared to the perceptions of laypeople. Research on knowledge about AI itself, showed that higher knowledge about Artificial Intelligence was related to lower risk perception, even though risks and benefits were balanced. We assessed for N = 750 participants (N= 544 laypeople, N = 66 medical experts, N = 11 AI experts, N= 129 (former) AI students) whether medical expertise, AI expertise, AI knowledge as well as self-reported medical expertise and AI knowledge (subjective) influence risk and benefit perceptions of medical AI applications. Result showed that, higher AI knowledge was associated with lower risk and higher benefit perception. That was also true in case of self-reported levels of AI knowledge. However, when comparing AI experts' risk and benefit perception with (former) AI students and laypeople, benefits were perceived as significantly lower. Furthermore, medical expertise and self-reported expertise were related to higher risk perception. These results reveal the intricate ways in which actual expertise and self-reported expertise shape the perception of medical AI, emphasizing the need for clear and tailored communication to bridge the gap between perceptions and the reality of these technologies.