Ethics and AI in the Healthcare Sector: Between Automation and Human Care
Artificial Intelligence is revolutionizing hospital wards, promising hyper-precise diagnoses and the dismantling of medical bureaucracy. But at what ethical cos
Medicine has always been a delicate balance between science and empathy. On one side, the rigor of clinical data, instrumental diagnostics, and pharmacology; on the other, the human touch, the understanding of suffering, and the therapeutic alliance between doctor and patient. Today, the entry of Artificial Intelligence into hospital wards threatens to forever alter this balance.
Predictive algorithms can identify a tumor in an X-ray months before it becomes visible to the human eye. Generative language models compile medical records in seconds, freeing doctors from the bureaucratic burden. However, delegating life-and-death decisions to a machine raises gigantic ethical questions: if an AI misdiagnoses, who is to blame? If an algorithm is trained on biased data, do we risk automating healthcare discrimination? And above all, are we running the risk of dehumanizing care?
In this in-depth analysis for the Scenarios and Reflections column, we will explore the complex relationship between Artificial Intelligence and medical ethics. We will analyze the guidelines drawn by the World Health Organization (WHO), the regulations of the European Union, the opinions of the Italian National Bioethics Committee, and the concrete use cases that demonstrate how AI must remain a tool in the service of humanity, and never its substitute.
1. The Global Context: The WHO Guidelines
Faced with an explosive and fragmented technological adoption, global health institutions have had to draw a clear ethical perimeter.
The fundamental document in this area is the report published by the WHO (World Health Organization) on Ethics and governance of artificial intelligence for health. The WHO established 6 cardinal principles (accompanied by 40 operational recommendations) that every developer and healthcare facility should respect:
- Protect human autonomy: The human being must remain in control of medical decisions and their own data.
- Promote human well-being and safety: AI must not cause harm (principle of non-maleficence).
- Ensure transparency and "explainability": A doctor must be able to understand why the algorithm suggested a particular diagnosis (avoiding the "Black Box" effect).
- Promote accountability: There must always be a clear chain of legal and moral responsibility in case of errors.
- Ensure inclusivity and equity: AI must not discriminate based on ethnicity, gender, or socioeconomic status.
- Promote responsive and sustainable AI: Systems must be ecologically sustainable and updatable over time.
These principles have also been promptly adopted by Italian medical associations, such as SIFO (Società Italiana di Farmacia Ospedaliera), which in its own document reiterated the importance of the ethics and governance of AI in the healthcare field outlined by the WHO, emphasizing how these recommendations should guide not only operators but also pharmaceutical companies and policymakers.
2. Automation vs. Human Care: The Risk of "De-skilling"
If AI is designed to support doctors, there is a concrete risk that it will end up making them lazy or devaluing their skills.
The National Bioethics Committee (CNB) and the National Committee for Biosafety, Biotechnology and Life Sciences (CNBBSV) have drafted a fundamental opinion in Italy entitled Artificial intelligence and medicine: ethical aspects. The document warns against a subtle phenomenon: the De-skilling of medical personnel. If young trainees get used to passively accepting diagnoses provided by AI without exercising their own clinical intuition and experience, they risk progressively losing the ability to formulate independent diagnoses. The machine should be a second opinion, not the sole decision-maker.
The doctor-patient relationship is intrinsically based on trust, empathy, and non-verbal communication, elements that the automation of routine cannot replicate.
The Challenge of Generative AI
With the arrival of Large Language Models (LLMs) in hospitals, the risks change. A briefing by the prestigious Hastings Center dedicated to Generative AI in Healthcare raises questions about the use of medical chatbots to interact with patients or draft clinical documentation. The ethics center recommends unwavering human supervision (Human Oversight) and watermarking practices to always make it clear to the patient when they are reading a text or receiving advice generated by a machine, in order to preserve the delicate relationship of trust with the medical institution.
3. Ethical Risks: Bias, Equity, and the Medical "Black Box"
The enthusiasm for algorithmic efficiency must not obscure the fact that Artificial Intelligence learns from past data. And if that data reflects centuries of healthcare inequalities, AI will do nothing but automate them.
Justice and Bias in Data
A rigorous review published in PMC (PubMed Central) analyzes the ethical issues of artificial intelligence in medicine. The study highlights how the risk of Bias is the most serious threat to the bioethical principle of justice. For example, many dermatological algorithms trained to recognize melanomas have been trained predominantly on images of Caucasian skin. Consequently, AI is dramatically less accurate in diagnosing tumors on dark skin, creating an unacceptable disparity in healthcare treatment.
Towards a New Governance
To mitigate these risks, an essay published in Nature points the way to shaping the future of AI in healthcare through ethics and governance. The scientific community demands absolute transparency (open source of medical datasets), accountability for developers, and an unwavering focus on global equity, so that AI does not become an elite tool available only in wealthy Western hospitals.
In Italy, the debate is fueled by specialized portals like InfoDottori, which in its analysis on ethics in artificial intelligence and healthcare robotics illustrates how facing the challenges of the future requires a continuous update of medical deontological codes, integrating technology without distorting the Hippocratic Oath.
4. Europe and the AI Act: The "High-Risk" Classification
While the academic world discusses principles, European legislators have turned ethics into law.
The European Commission has published rigorous guidelines on Artificial Intelligence in healthcare and digital care. With the final approval of the AI Act, Europe has classified almost all medical devices based on Artificial Intelligence as "High-Risk" systems.
What does this mean concretely? It means that before it can be used on a European patient, an AI software for diagnosis or triage must pass exhaustive tests to demonstrate the absence of racial or gender bias, must guarantee very high cybersecurity standards to protect sensitive health data (GDPR), and must include a mandatory human interruption mechanism (Kill Switch) in case the system starts producing anomalous results. Europe has chosen the path of ethical guarantee, even at the cost of marginally slowing down the market entry of new technologies compared to the deregulated American or Chinese model.
5. Practical Cases: From Psychiatry to the Elderly
Ethical implications become tangible when we observe the application of AI in contexts of extreme human vulnerability.
The Elderly and the Ethics of Surveillance
With the global aging of the population, assistive technologies are invading nursing homes. Predictive sensors and smart cameras analyze the gait of the elderly to prevent falls. However, as we explored in our in-depth analysis on AI and the Elderly: Active Aging or Digital Surveillance?, this technological assistance raises a huge ethical dilemma. Constantly monitoring a fragile individual guarantees their physical safety, but erodes their privacy and dignity. AI in this sector must be designed to support active aging and assist nurses (Human-in-the-loop), not to become a digital panopticon that replaces the warmth of a human visit with the coldness of a report on a tablet.
AI in Mental Health: The Therabot Case
Another minefield is that of algorithmic psychotherapy. Our special on AI and Anxiety Disorders: The Clinical Evidence of Therabot and Support Apps analyzes the use of chatbots based on Cognitive-Behavioral Therapy (CBT). In this context, the Stepped Care model applies: AI is ethically acceptable as a first-level tool to offer immediate support in moments of mild anxiety crisis, but it must never be presented to the patient as a substitute for a human therapist. Entrusting the resolution of deep trauma to an algorithm devoid of true empathy is not only clinically ineffective but morally dangerous.
Strategic Key Points
- The WHO's Warning: The 6 global principles establish that medical AI must protect patient autonomy and demonstrate transparency, equity, and legal accountability.
- The Bias Problem: Algorithms inherit human prejudices. If trained on unrepresentative data, AI diagnostic systems can discriminate against entire population groups based on ethnicity or gender.
- Medical De-skilling: The National Bioethics Committee warns that excessive dependence on AI risks atrophying the diagnostic intuition and clinical experience of young doctors.
- The European AI Act: The EU has classified AI in the medical field as "High-Risk", imposing strict controls on data quality and the obligation of human supervision before market entry.
- Care vs. Surveillance: In vulnerable sectors such as elderly care or mental health, technology must support the work of caregivers, not replace the human relationship or turn into oppressive digital surveillance.
FAQ: Ethics, Medicine, and Artificial Intelligence
1. Will Artificial Intelligence ever be able to replace doctors in the future? The unanimous consensus of institutions like the WHO and the academic community is "No." AI will replace tasks, not professions. It will replace transcribing reports or the visual analysis of X-rays, but interpreting the overall clinical picture, communicating dire diagnoses, and choosing therapy remain exclusively human prerogatives.
2. What happens if an AI makes a fatal error in the operating room? Who is to blame? Currently, jurisprudence and bioethical principles (including the AI Act) establish the principle of "Human Responsibility." The doctor who endorses the AI's suggestion (Human-in-the-loop) remains the final responsible party for the therapeutic decision. If the error is due to an intrinsic malfunction or a software bug, civil liability may fall on the company that developed the medical device.
3. What is the "Black Box" effect in medicine? Many Deep Learning algorithms are so complex that not even their programmers can explain exactly how the neural network arrived at a particular diagnostic conclusion. This is a huge ethical problem in medicine (Black Box): a doctor cannot administer chemotherapy just because "the computer says so," without understanding its clinical rationale. Hence the need imposed by Europe for "Explainability" (Explainable AI or XAI).
4. Is it ethical to use AI chatbots to treat depression or anxiety? It depends on the severity level. Guidelines indicate that the use of chatbots (like Therabot) is ethical if employed for daily well-being, mood monitoring, or as temporary support while waiting for an appointment. However, it is considered unethical and dangerous to entrust the autonomous care of serious psychiatric conditions (major depression, suicide risk) to an AI without the accompaniment of a human psychotherapist.
5. Is the health data I give to an AI-managed medical app safe? In Europe, health data is protected by Article 9 of the GDPR as "super-sensitive" data. Any medical application based on AI must guarantee the pseudonymization or anonymization of data before using it to train its models. However, the user must always carefully read the privacy policies, especially when using apps developed outside the European Union.
Conclusions: The Engineering of Empathy
The German philosopher and physician Karl Jaspers recalled that medicine is the most scientific of the humanities and the most humanistic of the sciences. Artificial Intelligence represents the pinnacle of the engineering and computational component of medicine, a tool capable of decoding the genome and uncovering invisible patterns.
But care is not just the absence of disease; it is a relational process in which one person entrusts themselves to another in a moment of extreme fragility. The algorithm can prescribe the statistically most effective drug, but it cannot hold a patient's hand before surgery, nor understand the weight of pain in a family's eyes.
The true ethical challenge of 2026 and the decades to come is not to reject Artificial Intelligence out of fear of progress, but to use it to automate everything that is mechanical, in order to give back to doctors the most precious asset that bureaucracy has stolen from them: the time to listen, look at, and care for human beings.
Bibliographic References and Sources
To ensure scientific, ethical, and institutional rigor, this article has drawn from the following primary sources:
- International Guidelines and Policy (EU/WHO):
- Academic Ethical Studies and Analyses: