AI Information & Limitations
AViTA Toolbox | AViTA Corporation
Last updated: 2026-10-28
Privacy Policy: https://sites.google.com/view/avita-toolbox
AViTA Toolbox | AViTA Corporation
Last updated: 2026-10-28
Privacy Policy: https://sites.google.com/view/avita-toolbox
AI-assisted features in AViTA Toolbox are intended to help present information and explore images. They are optional, and their availability may depend on the Application version, settings, and applicable service requirements. This page explains what the features do, what information is processed, how results should be understood, and their limitations.
Important: AI output is not a medical diagnosis or professional medical advice. An AI-generated description, enhanced photo, or arrow annotation must never be treated as proof that a medical condition exists or does not exist.
This feature can use Google Firebase AI Logic / Gemini to provide a plain-language interpretation of supported health measurements displayed in the Application, such as body temperature, blood pressure, heart rate, and SpO₂.
When you request this feature, the relevant measurement values and analysis instructions are transmitted to the third-party AI service. The response is generated from the information submitted and may not account for your medical history, symptoms, measurement conditions, or other important clinical context.
The otoscope image enhancement feature uses Apple Core ML and on-device super-resolution models to enhance or enlarge images captured through a compatible Wi-Fi otoscope. The Application may offer comparisons between original and AI-enhanced results.
This processing takes place on your device for the image enhancement function; it does not require sending the selected photo to a cloud AI service. Original and enhanced images may look different because super-resolution generates estimated details, not independently verified facts.
When enabled, the otoscope Cloud AI feature can send a photo you actively select to Google Gemini API for an AI-generated description of visible features. The service may also return suggested target coordinates that the Application uses to draw arrows and labels over the photo.
These descriptions and markers are generated estimates. They may be incomplete, inaccurate, or misplaced. They are not clinical findings, a diagnosis, or a substitute for assessment by a qualified professional.
Local AI Image Enhancement
Processing location: on the iPhone or iPad using Core ML.
Information used: the selected otoscope image.
Internet required for model inference: no, once the required models are present on the device.
Image sent to a third-party AI service for this enhancement: no.
Typical output: an enhanced or enlarged image for visual comparison.
Main limitation: newly generated detail may not accurately represent the original subject.
Cloud AI Image Analysis
Processing location: Google Gemini API servers.
Information sent: your selected image, analysis instructions, and the authentication data required for the API request.
Internet required: yes.
User-supplied API key: required in the current otoscope Cloud AI implementation.
Typical output: a text description and optional on-image arrows or labels.
Main limitations: potential factual errors, uncertain image localization, third-party data processing, and service-specific terms.
Health Measurement AI Interpretation
Processing location: Google Firebase AI Logic / Gemini services.
Information sent: relevant measurement values and analysis instructions when you request the feature.
Internet required: yes.
Typical output: a generated explanation of the provided values.
Main limitation: the model may not know the full circumstances of the measurement or the user's health.
Super-resolution models estimate image detail from patterns learned during model development. This may improve apparent sharpness or texture but cannot reveal information that the original capture did not reliably contain. AI processing may:
Introduce features or textures that were not present in the source image.
Remove or soften small details that were present.
Change apparent edges, color, contrast, or surface patterns.
Make compression artifacts appear more convincing.
Create images that look clearer without increasing the certainty of what is shown.
For a fair comparison, keep the original capture available. When assessing visual detail, rely on the original image and seek qualified professional advice where needed.
Cloud AI image analysis requires a user-initiated request. When you choose to analyze a photo, the Application sends that selected photo and related instructions directly to Google Gemini API through an encrypted HTTPS connection.
The Application does not upload all photos automatically simply because you connect the otoscope, record an image, open Files, or use Local AI. You must actively request Cloud AI analysis.
Before the first Cloud AI upload, the Application presents an information and consent prompt. Your approval may be remembered locally, but each new analysis remains an active choice. If you do not agree, you can use the local otoscope viewing, recording, and image-enhancement features without submitting images to Gemini.
The Application uses a Gemini API key supplied by you and stores that key in iOS Keychain. Removing the API key in Cloud AI settings prevents further requests using that key from your device and resets the saved upload consent. Removing the key does not erase earlier requests from Google's systems.
The Application's current Gemini Interactions API requests specify `store: false`. This opts out of storing the interaction as a reusable Interactions API object. It does not guarantee zero retention, zero logging, or zero use of data by Google.
Google's terms distinguish unpaid and paid services:
For certain unpaid services, submitted content and model outputs may be used to improve Google products and may be reviewed by humans. Google advises against submitting sensitive, confidential, or personal information to unpaid services.
For qualifying paid services, Google states that submitted prompts and responses are not used to improve its products, but limited processing and retention may still occur for security, abuse prevention, and legal compliance.
The treatment of a particular request depends on the Google project and service configuration, not simply on whether you entered your own API key. Otoscope images may contain sensitive health-related information. Do not send sensitive or personal information to a Google service when its terms or privacy protections do not permit that use.
Official references:
Gemini API Additional Terms: https://ai.google.dev/gemini-api/terms
Gemini Interactions API: https://ai.google.dev/gemini-api/docs/interactions-overview
Google Privacy Policy: https://policies.google.com/privacy
A Cloud AI response may:
Describe a normal visual feature as unusual, or overlook something important.
Misidentify a texture, color variation, reflection, shadow, or image artifact.
Give a confident explanation without sufficient supporting evidence.
Generate text that is incomplete, inconsistent, or factually wrong.
Draw an arrow or label at an incorrect position.
Fail to detect image blur, poor illumination, camera angle issues, or incomplete views.
Change its answer when the model version or analysis instructions change.
Annotations are visual aids displayed by the Application. They do not confirm that the highlighted location has any medical significance. Users should compare AI output with the unmodified photo and must not rely on an AI response for treatment decisions.
Cute Mode provides optional playful animations and background music intended to attract a child's attention while a parent or caregiver operates the otoscope.
Cute Mode:
Is controlled by the adult operating the Application.
Does not change the underlying Wi-Fi imaging or recording protocol.
Is not a child-operated account or a separate child-focused AI interface.
Does not, by itself, analyze photos or upload them to an AI provider.
Does not make Cloud AI appropriate for minors or remove service age restrictions.
Parents and caregivers should supervise the examination and use care when recording or sharing a child's image.
Google Gemini API terms impose age, permitted-use, geographic, and medical-use restrictions. The current terms prohibit using the service in clinical practice, to provide medical advice, or for activities requiring medical device regulatory clearance or approval, and restrict use through applications directed to or likely accessed by individuals under 18.
Accordingly, the existence of a Cloud AI option in the Application should not be interpreted as permission to use Gemini for diagnosis, treatment, regulated medical purposes, or children's medical-image analysis. Availability of the feature may be restricted or changed to satisfy provider terms and applicable law.
The Application's health-related information and AI-generated content are not emergency tools. If a person has pain, bleeding, fever, hearing changes, unusual readings, or other concerning symptoms, seek advice from a qualified healthcare professional. Contact local emergency services for emergencies.
You can choose whether to:
Use measurement AI interpretation, where available.
Use on-device AI enhancement for an otoscope photo.
Submit a selected image for Gemini Cloud AI analysis.
Store, share, export, or delete captured photos and videos.
Enter, replace, or remove your Gemini API key.
Turn Cute Mode animations and music on or off.
To stop further otoscope Cloud AI requests using a stored key, open Cloud AI settings and remove that key. To manage local images, use the Application's Files/media functions. To manage a copy exported to iOS Photos, use the Photos app. Earlier data sent to Google remains subject to Google's applicable retention and deletion practices.
Does taking an otoscope photo automatically upload it to Google?
No. Capturing and viewing a photo are local functions. A separate user action is required to request otoscope Cloud AI analysis.
Does Local AI send images to Gemini?
No. The otoscope Local AI image-enhancement function runs on the device using Core ML.
Can AI make an image more accurate?
Not necessarily. Enhancement may make the image appear sharper while generating inaccurate or artificial details.
Are AI arrows or labels medically verified?
No. They are machine-generated visual suggestions and may point to the wrong location.
Is a personal Gemini API key enough to ensure medical-image privacy?
No. Data use and retention depend on the applicable Google service tier and terms. Do not upload information that the selected service is not permitted or appropriate to process.
Does turning on Cute Mode let a child use AI?
No. Cute Mode consists of visual and audio effects for an adult-operated examination and does not authorize independent AI use by children.
Does removing my API key delete previous uploads from Google?
No. It prevents future requests from this device using that key and resets the local upload-consent setting, but provider-side records follow Google's own terms and controls.
For full information about data categories, third-party services, retention, security, user rights, and contact details, see the AViTA Toolbox Privacy Policy:
https://sites.google.com/view/avita-toolbox
For questions about these AI features or privacy matters, contact:
AViTA Corporation
Email: app_rd@avita.com.tw
© 2026 AViTA Corporation. All rights reserved.