Recent developments in medicine have drawn comparisons to scenarios found in books or science fiction films. Artificial intelligence (AI) is increasingly playing a role in health, transforming research, diagnosis, and patient support. However, its deployment raises crucial questions about ensuring that this technological revolution benefits everyone without exacerbating inequalities or sacrificing patient trust. A new podcast produced in collaboration with Amgen explores both the concrete benefits of AI and its ethical and practical challenges. AI is already transforming the care journey, from personalized treatment to identifying patients eligible for clinical trials. For example, the AIIPIK project, led by the startup Ospi and the Toulouse University Hospital with the support of Amgen France, identified 24 patient files eligible for a clinical trial, compared to only 6 in real conditions, while significantly reducing the analysis time for healthcare professionals. Marie Morice, Director of Innovation at Amgen France, notes this development. Many patients today discover the power of AI but question its relevance and the use of their data before perceiving the medical benefits. According to a study published in Nature in 2026, ChatGPT Health, although effective for simple questions, fails to recommend emergency care in critical situations. However, more than 230 million people consult it every week for medical advice. Solutions such as Gustave, the AI agent of Paperdoc, rely exclusively on validated French references (HAS, ANSM, scientific societies), guaranteeing scientifically reliable answers. AI raises important questions regarding the digital divide, which risks excluding certain patients, especially those less connected. "When it comes to conversational tools, the patient may find themselves facing an interface without mastering its codes," says Damien Dubois, a health communication consultant specializing in patient engagement. AI could then exacerbate inequalities in access to care rather than reduce them. Dehumanization of care is another obstacle, with the risk of healthcare professionals relying excessively on automated tools, at the expense of the human relationship and medical responsibility. For AI to be accepted, two conditions are essential: the quality of the data and transparency. "It is necessary to put in place real controls, verify that the data are reliable, comprehensive and up to date," insists Marie Morice. Without this, the risk is great of developing a lasting distrust toward these tools. Another challenge: accompanying patients through education. "How can we help them use these tools with the right references?" asks Marie Morice. It is necessary to learn to question the answers of a tool, identify reliable sources, and understand that an AI can make mistakes. The Illustrated Use Case Collection of AI in Health, published by France Assos Santé, is in line with this approach. Through five concrete examples, it shows how AI is already improving diagnoses, the organization of care, and the support of patients. "The goal is to speak with one voice and report upwards to a more institutional level so that the digital strategy on AI can support and highlight the solutions that are most virtuous," says Damien Dubois. AI will certainly never replace the care relationship or medical expertise. "It remains a tool to nourish reflection, a decision-making aid, but humans must retain control," recalls Damien Dubois. Marie Morice adds: "AI will likely be naturally integrated into daily tools, with increasingly concrete benefits for patients." The real challenge is not only technological. It is the collective ability to build an AI in health that is useful, understandable, accessible, and trustworthy. This increase in skills cannot be carried by a single actor. Patient associations, pharmaceutical laboratories, public institutions, and startups must move forward together. This is particularly the ambition of the Radar consortium, of which Amgen is a founding member, which brings together public and private actors around the reduction of diagnostic wandering in rare diseases through health data and AI.