AI processing summary
MedaStaré uses external AI services to provide conversational, styling, visual, voice, video and 3D-related features. The current production technical inventory identifies OpenAI, Google Gemini, Kling AI and Meshy as active AI providers.
Before an optional transfer of user data to an external AI provider where consent is required, MedaStaré presents an AI notice that identifies the provider, the relevant data categories, the purpose of processing, and available retention/training information, and provides a choice to continue or not proceed. If the user declines, the relevant AI request is blocked. Withdrawal is designed to stop future transfers that depend on that consent.
MedaStaré does not promise that every external AI provider offers zero data retention or zero model training. Provider-side retention and training depend on the provider, product plan, API configuration, contract and feature. A specific feature notice controls where it provides more current provider-specific information.
1. Scope
This AI Data Processing Notice supplements the MedaStaré Privacy Policy and applies when MedaStaré uses an artificial-intelligence or machine-learning provider to process user-selected information or account context for a requested feature.
It covers MedAI chat, AI-assisted style and wardrobe functions, selected visual-analysis functions, voice interaction, image-related processing, video generation, 3D generation and other AI-enabled features that identify this Notice or use the providers described below.
Not every provider or data category is used in every feature. MedaStaré routes only the information reasonably necessary for the selected function, subject to the consent and product controls described below.
2. Current AI providers
| Provider | Current MedaStaré role | Verified location / configuration note |
|---|---|---|
| OpenAI | Chat, text processing, text-to-speech, realtime voice and image-input processing where used. | Provider-managed API endpoints. MedaStaré does not publish a fixed processing country for OpenAI unless verified for the applicable account and product configuration. |
| Google Gemini | Chat, visual/image-related AI processing and text-to-speech where used. | Provider-managed endpoints. MedaStaré does not publish a fixed processing country for Gemini unless verified for the applicable account and product configuration. |
| Kling AI | AI video generation and related video-processing requests. | The current MedaStaré implementation uses a Singapore endpoint. |
| Meshy | AI-assisted 3D generation or related 3D asset processing. | Provider-side location and retention can depend on the Meshy plan and configuration. MedaStaré does not promise a fixed region or retention period unless verified for the active account. |
MedaStaré may add, remove or replace an AI provider as the product changes. Where a change materially affects the identity of the provider, the purpose, the type of personal data transferred or a legally required consent, MedaStaré will update the relevant notice and obtain a new choice where required.
3. Data that may be sent to AI providers
Depending on the feature the user chooses, an AI request may contain one or more of the following categories:
- Prompts, messages, conversation history and prior AI context needed for a multi-turn response.
- Style preferences, mood, occasion, color, fit, sizing or other personalization choices.
- Wardrobe information, garment descriptions, selected wardrobe images or saved-outfit context.
- Body-related styling information, measurements, proportions or silhouette context where the user has enabled a feature that needs it.
- Limited cycle-related context where the user has enabled that feature and the AI function requires the information.
- Images or other media deliberately selected for a visual, style, body, nail, beauty, generation, editing or similar AI feature.
- Voice/audio transmitted during an active voice interaction.
- Generated outputs or previous assistant responses where needed to continue a conversation or transform an existing output.
- Technical request information necessary to deliver, secure or troubleshoot the AI request.
The current technical inventory confirms that user preferences, wardrobe data, body data, limited cycle data and chat history can be transmitted to an AI provider when a feature requires them. The inventory does not establish that every category is sent to every provider or in every request.
Beauty-related information is not treated as universally transmitted merely because MedaStaré offers beauty features. It is sent only where the selected feature actually requires that information and the applicable transfer has been disclosed.
4. AI consent flow
For AI processing that requires user consent before an external transfer, the current production flow is designed to provide the following controls before the request is sent:
- A notice that data will be sent to an external AI provider.
- The provider name.
- The relevant categories of information to be transmitted.
- The purpose of the AI processing.
- Available information about provider retention and model-training practices.
- An Allow and Continue choice.
- A Not Now choice.
If the user chooses not to proceed, the relevant request is designed to be blocked at the backend and the unrelated parts of MedaStaré remain available.
The current consent implementation records consent state in MedaStaré's Firestore-based privacy records. The technical inventory confirms that consent version and timestamp are recorded. It does not confirm provider name, data categories, locale or app version as separate persisted audit fields in the current consent record.
Consent and related audit records may be retained as evidence of user choices, compliance and security. The current audit-log implementation does not apply an automatic expiration period to every consent audit record.
5. Data minimization and feature necessity
MedaStaré does not intentionally send all data in a user's account to every AI provider. The request payload should be limited to information reasonably necessary for the selected feature and the relevant conversation or generation context.
If a feature can operate without a sensitive category, that category should not be included merely because it exists in the account. Where technically appropriate, MedaStaré may use derived labels, summaries or limited context instead of sending a larger source record.
Users should avoid including unnecessary passwords, government identification numbers, payment-card details, medical records or other information unrelated to the requested AI feature.
6. MedAI chat and conversation context
MedaStaré stores AI responses and conversation history so that users can return to prior conversations and so that relevant conversation context can be included in later requests.
The current implementation does not offer a universal setting that disables chat history for all conversations. Users can delete individual chats where the product provides that control.
Under the current production configuration, prompts and chat history stored by MedaStaré do not have an automatic time-to-live expiration and may remain associated with the account until the user deletes the applicable chat, the account is deleted, or MedaStaré applies a shorter product or legal retention rule.
The current technical inventory states that prompts are not intentionally written into ordinary production runtime logs. Vercel runtime logs have an approximately one-day retention window under the current production plan, but server or security metadata may exist independently of chat-history storage.
7. Voice and realtime AI
MedaStaré currently uses the OpenAI Realtime API for supported realtime voice functionality.
When a voice session is started, microphone audio must be transmitted to the active voice provider so the request can be processed. Under MedaStaré's current implementation, raw voice audio is not intentionally stored by MedaStaré as a persistent account asset.
A transcript may be created and stored as part of the MedAI conversation. Under the current configuration, transcript retention follows the same general rule as other chat messages and does not have a universal automatic expiration period.
Provider-side handling of realtime audio is governed by the active provider's API terms, configuration and retention controls; MedaStaré does not represent that provider-side audio processing is zero-retention unless that setting has been specifically verified.
8. Image, video and 3D processing
MedaStaré may transmit user-selected media to an external AI provider when the user requests a feature that requires visual, video or 3D processing.
The current production inventory confirms that OpenAI can receive text and image inputs. OpenAI image generation/editing is not identified as active in that inventory. Google Gemini is used for chat and visual/image-related processing, Kling AI is used for video generation through the current Singapore endpoint, and Meshy is used for 3D-related processing.
Media sent for one requested feature is not, by that fact alone, authorized for unrelated advertising, profiling or another MedaStaré purpose.
Generated outputs may be saved to the MedaStaré account where the product offers a saved-output or history feature. If saved, the output is then subject to the MedaStaré Privacy Policy and Account and Data Deletion rules.
9. Sensitive and higher-risk inputs
Some AI features may involve information that is sensitive under applicable privacy law or sensitive because of context, including face/body images, body measurements, limited cycle-related context or private conversations.
MedaStaré's current AI features are not designed to identify a user by biometric identity or authenticate a user by face recognition. Visual analysis may detect appearance or body-related characteristics for the requested styling or beauty function without converting that function into biometric identification.
Where explicit consent or another enhanced legal basis is required for sensitive processing, MedaStaré will use the legally required basis before the relevant optional processing. Refusing such processing may make that specific feature unavailable but does not remove unrelated account functionality.
Cycle-related information is not used by MedaStaré for advertising.
10. MedaStaré retention
| AI-related category | Current MedaStaré retention approach |
|---|---|
| Prompts | No universal automatic TTL in the current production configuration. May remain until user deletion, account deletion or a shorter applicable rule. |
| Chat history and AI responses | No universal automatic TTL in the current production configuration. Individual chat deletion may be available. |
| Conversation context | Relevant prior context may be reused in later requests while the conversation remains available. |
| Raw voice audio stored by MedaStaré | Not intentionally retained as a persistent account asset in the current OpenAI Realtime implementation. |
| Voice transcript | Follows the same general retention rule as other stored chat messages. |
| Runtime logs | Approximately one day for Vercel runtime logs under the current production plan, separate from stored chat history. |
| Consent audit records | May be retained to demonstrate the user's choice and compliance. The current audit log does not apply an automatic expiration period to every entry. |
Account deletion removes AI-related MedaStaré account data covered by the deletion workflow, subject to the limited security, audit, transaction and legal exceptions described in the Account and Data Deletion page.
11. Provider retention and model training
External-provider retention is different from MedaStaré's own retention. The provider may temporarily process or retain request data under its API/business terms, security controls, abuse-monitoring systems, selected features or account configuration.
The signed MedaStaré technical inventory did not verify one fixed provider-retention period or one universal training configuration for OpenAI, Gemini, Kling and Meshy. MedaStaré therefore does not make a blanket contractual promise that every provider operates with zero data retention or zero model training.
As of this Notice, OpenAI publicly states that content submitted through its API/business offerings is not used to train its models by default unless the customer affirmatively opts in to data sharing. Google states that prompts and responses submitted through paid Gemini API services are not used to improve Google products, while provider-side logging or retention can still occur depending on the Gemini feature and configuration. Meshy retention and training practices can vary by product plan. Kling provider-side retention and training settings must be evaluated under the active Kling terms and configuration.
A provider's public policy can change independently of MedaStaré. Where MedaStaré has contractually enabled a stronger setting such as zero-data-retention or enterprise controls, that stronger setting will be stated only after it has been verified for the applicable production account.
MedaStaré does not use private account content to train a generally available MedaStaré model without a separate affirmative choice.
12. Processing locations and international transfers
MedaStaré is a United States company and its core application infrastructure is primarily hosted in the United States. AI processing can occur in other countries depending on the provider.
The current production implementation verifies a Singapore endpoint for Kling AI. OpenAI and Google Gemini use provider-managed general endpoints under the current configuration, and the technical inventory does not verify one fixed processing country for those providers. Meshy's processing location depends on the active service and plan and is not represented here as a fixed region unless separately verified.
Where applicable law requires a cross-border transfer mechanism, MedaStaré will use an available lawful mechanism appropriate to the relevant provider and transfer, such as contractual safeguards, an adequacy mechanism, recognized certification or another legally permitted safeguard.
13. Human access, safety and moderation
AI requests are primarily processed automatically. MedaStaré personnel do not receive unrestricted access to private AI content merely because the content has been processed by an AI feature.
Authorized MedaStaré personnel may access limited AI-related content or metadata when reasonably necessary for user-requested support, trust and safety, abuse or fraud investigation, security, incident response, debugging, legal compliance or the establishment, exercise or defense of legal rights.
External providers may also use automated safety systems and, where allowed by their applicable business/API terms and configuration, limited human review for abuse, security or service integrity. MedaStaré does not promise that an external provider will never conduct human review unless that restriction has been contractually and technically verified.
Where user-generated media is subject to community moderation, separate automated moderation services and human review may be used before publication or enforcement. A provider's automated safety result does not necessarily constitute MedaStaré's final moderation decision.
14. AI outputs and automated decisions
AI outputs can be inaccurate, incomplete, biased, outdated or unsuitable for a particular user or situation. MedaStaré's fashion, style, wardrobe, beauty and conversational outputs are informational and should be reviewed with independent judgment.
MedaStaré does not use fashion or beauty AI alone to make a decision that produces legal or similarly significant effects concerning a user.
Automated signals may be used for moderation, abuse prevention, eligibility, ranking or challenge administration. Where a MedArena or reward feature uses automated scoring or selection, the applicable Official Rules and Terms govern eligibility, review, disqualification and any human oversight.
AI-generated content remains subject to MedaStaré's Terms of Use, intellectual-property rules, safety rules and any applicable third-party rights.
15. Withdrawal, deletion and provider-side records
Where consent is the basis for an optional AI transfer, withdrawing consent is designed to stop future transfers that depend on that consent. Withdrawal does not make earlier lawful processing unlawful and does not automatically erase processing that already occurred at an external provider.
Users may delete individual chats where that control is available and may delete their MedaStaré account through the account-deletion flow. Account deletion removes AI-related data from MedaStaré-controlled active systems as described in the Account and Data Deletion page.
Provider-side deletion is subject to the provider's technical capabilities, API controls, retention configuration, security requirements and legal obligations. MedaStaré will use available provider-side deletion mechanisms where required and technically supported but cannot promise deletion of a provider record that the provider is independently required or permitted by law to retain.
16. Provider or feature changes
MedaStaré may change AI models or providers for quality, availability, cost, security, legal or product reasons. A model version change that does not materially change the disclosed processing may not require a new notice.
If a change materially changes the provider, purpose, data category, sensitive-data use or consent requirement, MedaStaré will update this Notice or the feature-specific notice and obtain any additional consent required by law before the new processing begins.
17. Questions and privacy rights
Questions about AI processing, consent, data retention or a particular provider may be sent to support@medastare.com.
Privacy rights, including applicable rights of access, correction, deletion, objection, withdrawal of consent or portability, are described in the Privacy Policy and may be submitted through the Privacy Choices page.
Please do not send passwords, authentication tokens, complete payment-card information, unnecessary identity documents, medical records or sensitive images by ordinary email.