Terms of Use
Terms designed for clinical intelligence that supports professionals without displacing clinical judgment, the care relationship, or the rules that apply to each deployment.
The operating agreement in plain language
- Use Sofya to organize evidence, document care, and support clinical reasoning.
- A qualified professional must review every output before relying on it or adding it to a record.
- Use identifiable patient data only in an expressly authorized institutional environment.
- Do not use the public experience for emergencies or as a direct-to-patient diagnosis service.
Agreement and scope
These Terms govern access to sofya.med, its public interactive experience, and any Sofya service that links to these Terms. “Sofya,” “we,” and “us” refer to the Sofya entity providing the applicable service. By using the service, you agree to these Terms and confirm that you have authority to do so.
A signed order form, enterprise agreement, business associate agreement, data processing agreement, research protocol, or other institutional contract controls if it conflicts with these public Terms. The intended use and regulatory status of a particular deployment are defined by its product documentation and contract, not by a marketing page or disclaimer.
Service provider and corporate identity
The applicable Sofya provider is the entity identified in the order form or institutional agreement. For the public site and direct relationships in Brazil, the relevant entity is Sofya Tecnologia Ltda. · CNPJ 46.163.535/0001-84 · Rua Padre Adelino, 2074, 12º andar, conjunto 121, Quarta Parada, São Paulo, SP 03303-000, Brazil.
For direct relationships in the United States, the relevant entity is Sofya AI Inc. · Florida document P25000007209 · 2 S Biscayne Blvd, Suite 2450, Miami, FL 33131, United States. A contract may identify another affiliated provider for a particular territory or service, and that identification controls for the contracted relationship.
Clinical purpose and professional control
Sofya may organize clinical information, surface evidence, suggest possibilities, support documentation, and help a care team follow a case over time. These capabilities are designed to augment professional work.
Sofya does not independently practice medicine or issue a final diagnosis. An output becomes part of a diagnosis, prescription, treatment, prognosis, referral, discharge instruction, or medical record only after a qualified professional reviews and adopts it. The responsible professional remains in control of the decision and the communication with the patient.
Review is required before clinical use
AI-generated content can be incomplete, outdated, imprecise, or wrong. Before relying on an output, the user must compare it with the patient’s presentation and record, confirm material facts, check primary sources when appropriate, and apply current evidence, local protocols, contraindications, allergies, interactions, dose limits, and the user’s own professional judgment.
Do not copy an output into a medical record without review. Preserve uncertainty, provenance, clinically relevant limitations, and the distinction between observed facts, patient reports, and generated inferences. If an output conflicts with the case, applicable guidance, or professional judgment, disregard it and use the established clinical escalation path.
Ethical AI in the care relationship
Responsible use keeps patients and health professionals at the center. The person remains more important than the prediction, and efficiency must not weaken consent, dignity, equity, confidentiality, or access to human care. The patient-professional relationship remains sovereign in clinical decisions: Sofya may strengthen that relationship, but must not invisibly mediate, assume, or replace it.
- Human autonomy: keep meaningful professional oversight and never coerce a patient or professional to accept an AI suggestion.
- Transparency: provide understandable notice when AI materially participates, the task it performs, its clinically relevant basis, provenance, and limits, and who remains responsible, subject to the professional’s and institution’s legal duties.
- Patient communication: diagnoses, prognoses, and treatment decisions must be communicated with appropriate human mediation; do not delegate difficult clinical conversations to the system.
- Sovereign care relationship: preserve the professional’s independent judgment and the patient’s values, preferences, questions, and right to human dialogue; the technology must remain support for that relationship.
- Patient choice and access: where law, institutional policy, or the nature of the workflow requires it, offer an understandable explanation and a meaningful way to seek human review or decline AI participation without improper loss of care.
- Bias and underserved populations: identify, measure, mitigate, monitor, and correct bias, with specific attention to historically underserved populations, local languages, disability, low-resource settings, and gaps in representation; investigate unexpected disparities instead of treating them as noise.
- Democratization with evidence: use the technology to broaden access to excellent clinical reasoning and help raise the standard of care globally, while respecting local evidence, culture, regulation, resources, and validated intended use.
- Safety and accountability: use the service only for validated and authorized workflows, report material errors, and maintain a path for review, correction, and contesting consequential outputs.
- Consequential decisions: do not use a Sofya output as the sole basis to deny or materially restrict care, benefits, coverage, employment, credentialing, or professional discipline where meaningful human review is required.
- Professional integrity: do not use automation to conceal authorship, fabricate evidence, or create a record that implies an examination or decision that did not occur.
- Professional agency: give health professionals appropriate information, training, and a way to question, override, or report the system; do not use Sofya as undisclosed surveillance of professional performance.
Patient information and authorized environments
You must have the authority, professional basis, and notices or consents required to enter patient information. Follow minimum-necessary principles and your institution’s policies. The public no-account composer is intended for synthetic or properly de-identified cases unless Sofya has expressly authorized your organization to use identifiable data there.
Audio and documents selected in the public composer are transmitted for transcription or extraction and may require transient processing outside the browser. The experience is designed not to create a persistent medical record, and clinical content is kept outside marketing analytics. Details appear in the Privacy Policy.
Emergencies and direct patient use
The public site is not an emergency service and must not be used to obtain time-critical assistance. In an emergency, contact the local emergency service or follow the institution’s emergency protocol.
Unless a deployment is expressly designed, authorized, and labeled for patient use, Sofya is a professional support tool and not a substitute for a consultation with a licensed clinician. Patients should not start, stop, or change treatment based only on generated content.
Brazilian clinical use
In Brazil, use must remain within the professional’s legal scope and the institution’s governance. Consistent with CFM Resolution 2,454/2026, AI supports clinical practice while the physician retains final responsibility for clinical, diagnostic, therapeutic, and prognostic decisions and critically reviews the result against the case, evidence, and good practice.
A software function’s regulatory treatment depends on its intended purpose and actual operation. Nothing in these Terms represents that every Sofya configuration has the same classification or authorization. Use only the features, indications, and environment approved for the applicable deployment.
United States clinical use
In the United States, users must act within applicable licensure, scope-of-practice, institutional, payer, federal, and state requirements. Where Sofya handles protected health information on behalf of a HIPAA covered entity or business associate, the applicable business associate agreement and institutional contract govern that processing. Do not submit PHI through a public or consumer-facing experience that is not expressly covered by those agreements.
FDA treatment of clinical decision support depends on the software function, intended user, intended use, and whether the health professional can independently review the basis for a recommendation. Access to Sofya does not by itself state FDA clearance, approval, or non-device status. Use each deployment only within its documented intended use and applicable regulatory permissions.
Permitted use and prohibited conduct
You may use the service for lawful clinical, operational, educational, research, evaluation, or institutional purposes within the access and rights granted to you. You are responsible for the legality and accuracy of what you submit and for actions taken from the output.
- Do not submit data you are not authorized to process or attempt to identify a de-identified person.
- Do not use the service to discriminate unlawfully, manipulate a patient, impersonate a professional, fabricate care, or evade required human review.
- Do not probe, disrupt, overload, bypass security, introduce malicious code, or access another person’s organization, patient, encounter, or credentials.
- Do not reverse engineer or extract models, prompts, software, or protected content except where a non-waivable law expressly permits it.
- Do not present generated content as independently verified, as Sofya’s medical opinion, or as a regulatory authorization.
Content, ownership, and feedback
You retain the rights you have in content you lawfully submit. You grant Sofya the limited rights needed to process that content, provide the requested function, protect the service, and meet legal obligations. Institutional agreements may define additional processing instructions.
Sofya and its licensors retain rights in the service, software, interfaces, models, documentation, and brand. If you provide product feedback, Sofya may use it without identifying a patient or disclosing your confidential information.
Availability, changes, and suspension
We may improve, limit, or discontinue public features and may suspend access when reasonably necessary for security, legal compliance, patient safety, service integrity, or breach of these Terms. Clinical organizations should maintain appropriate downtime and fallback procedures rather than make care depend on uninterrupted availability.
Risk allocation and mandatory rights
To the extent permitted by law, public and evaluation features are provided without a promise that every output will be complete or suitable for a specific patient. Sofya does not exclude liability that cannot lawfully be excluded, including mandatory consumer rights or liability arising from fraud, willful misconduct, or other non-waivable duties.
For contracted services, warranties, service levels, indemnities, liability limits, governing law, and dispute procedures are those in the applicable institutional agreement. Otherwise, these Terms are governed by the laws applicable to the Sofya entity providing the service, while preserving mandatory protections that apply to you.
Updates and contact
We may update these Terms to reflect changes in the service, law, or clinical governance. We will post the revised effective date and provide additional notice when required. Continued use after an effective update means acceptance where permitted by law.
Questions about these Terms, responsible use, or the correct institutional environment may be sent to privacy@sofya.ai. Do not include patient information that is not necessary for the request.
