BANANA TOUCH
KOR

AI Use Notice

Effective date: October 1, 2026 · Last revised: October 1, 2026

Operated by Bananatouch Inc. · contact@bananatouch.com

Reference translation. This English version is provided for convenience only. If it differs from the Korean original, the Korean original prevails.
Key notice. AI-generated record recognition and structuring results, summaries, classifications and suggestions may contain errors, omissions or bias. They do not replace the professional judgment of staff (medical personnel, care workers, etc.) who have checked the source data and the condition of the care recipient, and must not be used as the sole basis for diagnosis, treatment, prescription, emergency decisions or care plans. In an emergency, call 119 or contact the nearest emergency medical facility immediately.
Contents
  1. 1. Scope
  2. 2. Disclosure of AI Use and Results
  3. 3. Staff Review and Human Oversight
  4. 4. Data Input Principles
  5. 5. AI Inputs/Outputs and External Providers
  6. 6. Automated Decisions and Objections
  7. 7. Prohibited Uses
  8. 8. Reporting Errors and Safety Issues
  9. 9. Related Documents

1. Scope

This notice explains the general principles of the AI-based features that Bananatouch Inc. (the “Company”) introduces on its website or provides under separate agreements, including AI record recognition (AI-OCR), item-by-item structuring, automatic masking of identifying information, summarization, classification, suggestions, chatbots, AI agents and reports. The specific features, intended use, scope of data processing, and whether a feature qualifies as a medical device or digital medical product are determined separately in each product screen, order form and agreement with the customer organization.

The record recognition demo on the website is an example based on the actual recognition screen, using samples created with fictitious information. It does not accept users’ documents.

2. Disclosure of AI Use and Results

  1. Features in which AI substantially contributes to generating a result are labeled so that users can easily recognize them.
  2. AI results are affected by the accuracy and completeness of the input data, the scope of model training and technical limitations.
  3. Even when a generated summary or suggestion reads naturally and confidently, it may be inaccurate or omit important context.
  4. Verifiable information about AI results, such as the source, time of generation and model, is provided within the scope of each feature and agreement.

3. Staff Review and Human Oversight

  1. Before using an AI result, the responsible staff member (medical personnel, medical technicians, care workers, social workers or other personnel designated by the organization) must personally check it against the original records, test results, the current condition of the care recipient and other clinical and care information.
  2. The service does not replace diagnosis, prescription or treatment planning by medical personnel, care planning by the organization, or explanations and consent required by law.
  3. Where an AI result may affect a person’s rights or safety, or an important medical or care decision, an authorized person must review it and make the final decision.
  4. The product must not be used outside its approval, certification or intended use. The applicable status should be checked in the guidance for each feature.
  5. Record recognition and structuring results are confirmed by staff after comparing them side by side with the original and correcting any errors.

4. Data Input Principles

  1. Actual personal information, health information, care records or medical records of care recipients must not be entered on the public website, in the consultation or brochure request forms, or in public demos without separate safeguards.
  2. In services for customer organizations, only information that the customer organization is lawfully authorized to process may be entered, and only to the minimum extent necessary for the task.
  3. Unique identifiers such as resident registration numbers are not entered unless a legal basis and separate safeguards have been confirmed.
  4. Information about users or third parties must not be uploaded without authorization or reused for other purposes.

5. AI Inputs/Outputs and External Providers

  1. Consultation and brochure request information submitted on the public website is not used as input or training data for AI models.
  2. Where AI features for customer organizations store text, voice, documents, images or generated results, or transmit them to an external AI provider, the purpose, items, provider, country of processing, retention period and legal basis are disclosed to the customer organization and users in a separate document.
  3. In services for customer organizations, identifying information written on records (name, resident registration number, contact details, etc.) is automatically detected and stored in masked form, and within the Company only the Chief Privacy Officer can view the original information.
  4. The Company uses records to improve and research AI models only where the customer organization has agreed in its agreement, and only after pseudonymizing them. Pseudonymized information cannot be used to identify a specific individual; additional information is stored separately, and no attempt is made to re-identify individuals. If the organization does not agree, records are not used for training.
  5. Whether external AI providers use data for training, and their retention settings, are confirmed in the agreement and operating environment, and any change is notified.

6. Automated Decisions and Objections

If the Company provides a feature in which AI uses personal information to make, in a fully automated manner, a decision that significantly affects a person’s rights or obligations, it will separately explain, before the feature is used, the applicable criteria, procedure, data used and impact, and how to refuse, request an explanation or obtain human review. A simple recommendation or summary for reference is not by itself presumed to be such an automated decision.

7. Prohibited Uses

  • Relying solely on AI responses to detect or report emergencies
  • Making final decisions on diagnosis, prescription, procedures, care plans or care recipient priority based solely on AI results
  • Presenting AI results to care recipients or their families as confirmed facts reviewed by staff
  • Using the features for discrimination, unlawful surveillance, re-identification or unauthorized inference of personal information
  • Circumventing or removing safeguards, access controls or result labels

8. Reporting Errors and Safety Issues

If you find a result that may be inaccurate or dangerous, an exposure of personal information, or a missing AI-use label, stop before using the result in actual care or treatment and notify contact@bananatouch.com, keeping any information that could identify a care recipient to a minimum. The Company will review the report and, where necessary, correct the result, restrict the feature, arrange human review or follow its security incident response procedure.

— This notice applies from October 1, 2026. —