Manipulated news images are as old as the medium; however, the use of in-house AI-generated genre images among legacy media raises a new set of ethical questions, as these potentially contribute to a broad distrust in all types of imagery. While most previous research has focused on journalism practices, which have been extensively scrutinized, this study identifies key challenges influencing how legacy media relate to the ethics of utilizing AI technology in their in-house image production in the early stages of this technology. The method of this study was semi-structured interviews with Swedish legacy media journalists. The theoretical framework is based on the concepts of applied journalism ethics and the concepts of credibility and accountability, as the ethical guide serves as the foundation for the legacy media’s norms and values in shaping the professional standards of visual journalism. The results show that due to the rapid development of AI, many Swedish legacy media outlets generate in-house AI visuals; however, most do not utilize photorealistic visuals in their news production, owing to ethical considerations. These considerations depend on visual competency and affect authenticity and are based on criteria emphasizing mainly transparency and truthfulness, and less on bias and copyright.