CFM Resolution 2,454/2026: AI in medicine has rules

Resolution 2,454/2026, from the Federal Council of Medicine (CFM), is the rule that regulates the use of artificial intelligence in Brazilian medicine. Published on February 27, 2026 and in force since August 26, it starts from a simple idea: AI is a support tool, and the final decision always belongs to the doctor.

What happened

The CFM approved the resolution on February 11, 2026, after about 18 months of work by a commission dedicated to the subject, and published it in the Diário Oficial da União (federal official gazette) on the 27th. The text gave hospitals, clinics and doctors 180 days to adapt, a deadline that ended on August 26.

The central points, according to the council itself and coverage by Correio Braziliense and Agência Brasil:

• The doctor decides. AI can suggest, calculate and rank hypotheses, but diagnosis, prognosis and therapeutic decisions are the doctor's responsibility. Doctors are not obliged to automatically follow a system's recommendation and may refuse a tool they consider inadequate or not validated.
• Communication is human. The rule prohibits delegating to AI the communication of diagnoses, prognoses or therapeutic decisions to the patient.
• The patient knows. Patients have the right to be informed, clearly, whenever AI is used in their care, and may refuse its use.
• Classified risk. Systems used or developed by medical institutions undergo an algorithmic risk assessment: low, medium, high or unacceptable.
• Records and data. The use of AI as support must be noted in the medical record, and systems must comply with the General Data Protection Law (LGPD) and medical confidentiality.
• Protection for the diligent doctor. Professionals are not liable for failures attributable exclusively to the system when they show critical, diligent and ethical use of the tool.

Why it matters

AI is already in doctors' offices. The TIC Saúde survey, by Cetic.br, released in May 2026, shows that 18% of the country's health facilities used some form of artificial intelligence in 2025, a figure that rises to 31% among those with more than 50 beds. Most of the use is still operational, such as organizing clinical and administrative processes.

Before the resolution, each service decided on its own how to deal with this. Now there is a common vocabulary: support, supervision, transparency, risk. For the patient, the most concrete change is the right to know when a machine took part in their care. For those developing health technology, the rule draws a clear line between what is an assistant and what would be a replacement for the doctor, and only the former is allowed.

In Bairo's view

For Bairo Leandro Gonzalez Martinez, the resolution puts into regulation what he defends as a vision: AI is an extension of the human, not its substitute. In his reading, a technology that organizes the patient's history, translates and summarizes gives the doctor time back for what only they do, which is to listen, examine and decide together with the person.

That is why, in the health ventures he founded, he speaks of a clinical assistant and never of a machine doctor. At Hospital Triage, the assistant talks with the person before the consultation and hands everything over, organized, to a doctor or nurse, who validates and decides. At Xperienc Global Labs, the teleconsultation with a specialist from another country is a clinical second opinion, not a diagnostic report. Bairo sees in this design a way to broaden access without taking the professional out of the center.

He also recalls that technology does not replace presence. For the pain that does not show up in a test, there are the pages on pain, faith and reconnection and the path of when to seek professional help.

Where this meets the ecosystem

Hospital Triage was conceived from the start with human validation at every stage. The assistant does not give a diagnosis and does not show a severity classification to the patient; when faced with a warning sign, it stops the conversation and points to 192 (Brazil's emergency ambulance number). The system exists as a prototype in a test environment and does not yet operate commercially.

Xperienc Global Labs is developing the Global Language translator and teleconsultation with a clinical assistant. Both projects were born with CLAIN, the ecosystem's AI agent orchestrator. When they reach operation in Brazil, they will have to go through the risk assessment the resolution requires.

Sources

• Federal Council of Medicine, "CFM normatiza uso da IA na medicina" (CFM regulates the use of AI in medicine), Feb. 2026: portal.cfm.org.br
• Federal Council of Medicine, "Regras sobre uso da IA na medicina entram em vigor nesta quarta-feira, 26 de agosto" (Rules on the use of AI in medicine take effect this Wednesday, August 26), Aug. 2026: portal.cfm.org.br
• Correio Braziliense, "CFM publica resolução que regulamenta uso de IA na medicina" (CFM publishes resolution regulating the use of AI in medicine), 2/27/2026: correiobraziliense.com.br
• Agência Brasil (Radioagência Nacional), "Médicos têm até 26 de agosto para se adaptar sobre uso de IA" (Doctors have until August 26 to adapt to the use of AI), 5/14/2026: agenciabrasil.ebc.com.br
• Cetic.br, TIC Saúde 2025, released on 5/12/2026: cetic.br

Read also

• CFM Resolution 2,314/2022: the telemedicine rules
• AI triage in the emergency room: what the studies say
• AI, an extension of you