Consent and Telling Patients That AI Was Involved

Patients have a right to know when artificial intelligence influences their clinical care. This right arises from ethical principles of respect for autonomy and informed consent, and increasingly from regulatory requirements. Yet many healthcare settings have not clarified whether, how and when to inform patients that an AI system played a role in their diagnosis, treatment plan or risk assessment.

The EU AI Act requires transparency when AI makes decisions affecting individuals. Article 50, in force since 2 August 2026, establishes transparency requirements for high-risk AI. These obligations extend beyond healthcare to other sectors, but many healthcare AI applications qualify as high-risk under the regulation. The law states that individuals must be informed when AI is used in ways that significantly affect them. This requirement is separate from and additional to any consent specific to a particular tool.

In practice, transparency means patients can understand that AI was involved and what role it played. It does not necessarily mean every patient needs detailed technical knowledge of how the algorithm works. A patient does not need to understand backpropagation or training data to understand that a diagnostic AI flagged a finding, which a clinician then reviewed and confirmed. Transparency means clarity, not unnecessary technical detail.

The timing and method of disclosure matter. Some organisations have chosen to inform all patients broadly that “AI-assisted diagnostic tools may be used as part of care” through general patient information leaflets. This approach is transparent but generic. Other organisations disclose specifically when a particular AI system was used in a patient’s care. Both approaches can be appropriate, depending on your setting and the extent to which AI is embedded in routine care.

Patients may have opinions about AI use in their care. Some welcome the use of diagnostic assistance tools because they understand these tools help clinicians make accurate diagnoses. Others may have concerns about privacy, or they may prefer care without algorithmic involvement even if it delays diagnosis. Patients may ask why an AI recommendation was overridden or trust it when clinicians questioned it. These conversations are part of good clinical practice.

Your own understanding of the AI system matters when speaking with patients. If a patient asks “how confident is the AI that I have this condition”, you need to know whether you are telling them about the system’s confidence scores (which may be high but are not the same as clinical probability) or your own clinical assessment (which is the one that actually matters). If a patient asks “what data was the system trained on”, you should know whether it was trained on patients like them or on a different population. These details affect what you can honestly tell patients.

The EU AI Act Article 50, in force since 2 August 2026, establishes transparency requirements for high-risk AI systems. These legal obligations exist for systems that are classified as high-risk under the regulation. Many healthcare AI applications will fall within this classification. This means transparency is not merely good practice, it is a legal requirement. Your organisation should understand which systems are high-risk and ensure compliance with transparency requirements.

Some patients will ask whether they can opt out of AI use in their care. Your organisation should have a clear policy on this. Can patients decline AI-assisted diagnosis? For some applications, probably yes. If a patient declines a diagnostic AI, can a clinician examine the images and provide opinion without the algorithm’s input? Usually. For other applications like ambient documentation or background analytics, opting out might be impractical. These policy decisions should be made at an organisational level, not left for individual clinicians to navigate.

Transparency serves patients by allowing them to make informed decisions about their care. It serves clinicians by reducing legal and ethical liability. It serves organisations by demonstrating compliance with regulation. These interests align when transparency is done well.

Summary

Patients have both ethical and legal rights to transparency about AI use in their care. The EU AI Act Article 50 requires transparency for high-risk AI systems. Transparency means clarity about when and how AI was involved, not necessarily technical detail. Your organisation should have clear policies about disclosure timing, method and what patients can do if they have concerns or wish to decline AI involvement.