CFM Resolution 2.454/2026, which takes effect in August 2026, is not only about risk classification. It establishes explicit patient rights regarding AI: to be informed when the technology is used in a relevant way in their care, to know which model is involved, which data is shared and what the limits are — and, if they wish, to refuse. A doctor who uses AI as support must record it in the medical record and respect the patient's autonomy. A clinic that automates confirmations and rescheduling needs to know where this rule applies and where it does not.

What the resolution requires in terms of transparency

When an AI system is used to support a technical decision — triage, protocol suggestions, exam analysis — the patient has the right to be informed. The rule requires clear information about the model used, its limits, the data shared and the risks. The use must be recorded in the medical record. Communicating diagnoses or therapeutic management remains prohibited for AI; that responsibility belongs to the doctor.

What changes in practice: if the clinic uses software that suggests diagnoses or prioritizes cases, the patient needs to know. If the clinic uses an AI agent that only confirms appointments and reschedules times over WhatsApp, the resolution classifies that as low risk — administrative functions with no direct impact on clinical decisions. The line sits at the border between “decision support” and “operational automation.”

The right to refuse and what it implies

The patient can refuse the use of AI in their care. The resolution is explicit. The doctor must respect that autonomy. For the clinic, this means: when AI touches a clinical decision, the flow must allow for the patient to opt for care without AI. In purely administrative tools — confirmations, reminders, rescheduling — refusal does not apply in the same way, because no clinical decision is involved.

The difference is important. A patient cannot “refuse” the clinic sending an appointment reminder over WhatsApp. They can ask not to receive it, but that is a channel preference, not a refusal of AI in a medical decision. A patient who will be assessed by a triage or diagnostic suggestion system, however, can refuse and demand an exclusively human assessment.

Where the small clinic fits in

Most clinics and private practices do not use AI for triage or diagnosis. They use it for scheduling, confirmation and communication. In those cases, the resolution does not require the patient to be told “which AI model” is in use — because there is no model making a clinical decision. The agent that confirms appointments and reschedules times runs defined flows; it does not suggest management or prioritize cases by severity.

This does not excuse the clinic from good practices: a clear privacy policy, consent for data use, LGPD compliance up to date. The CFM resolution does not replace the LGPD; it complements it, focusing on what is specific to medical practice and to the use of AI with clinical impact.

What to do in practice

If the clinic uses or plans to use AI that supports decisions (triage, protocol suggestions, exam analysis): inform the patient, document it in the medical record, and make sure the flow allows refusal. If the clinic uses only administrative automation — confirmations, reminders, rescheduling — the resolution does not require each patient to be informed about “which AI” is in use. What remains mandatory is data protection and transparency about how data is handled.

For the clinic manager evaluating tools: ask the vendor whether the system makes or influences clinical decisions. If the answer is no — if it is only front-desk automation — you are in low-risk territory. If the answer is yes, the patient has the right to know and to refuse.

The timing is for preparation

The resolution takes effect 180 days after publication (August 2026). Those who already use AI in clinical decisions need to adapt. Those who use only administrative automation are already aligned with the low-risk classification — what the resolution reinforces is the patient's right when AI really matters for care.

Want to see how it works in your clinic? Try it free for 10 days at simplafy.com.br/saude — no credit card.

References used to produce this content: