Privacy Policy.

Dafdef, inc., doing business as Interfaces (“Interfaces,” “we,” “us,” or “our”), builds new ways for people to communicate with and use AI. This Privacy Policy explains how we handle information across every Interfaces product and service, including our applications, websites, AI models and agents, Silent Speech and voice features, developer tools and APIs, hardware we may offer, and research previews (collectively, the “Service”).

This is our one company-wide privacy policy. Feature screens may explain a collection at the moment it happens, but they do not create separate product privacy policies. Because the Service is currently a research preview, account users must also accept our Research Agreement.

1. Roles and scope

Interfaces is generally the controller of information from our websites, account holders, Personal experience, and our own research. When a developer sends end-user data through an Interfaces API, Interfaces may instead process that data for the developer under the developer’s instructions. Developers remain responsible for their end-user notices, permissions, legal basis, and requests.

Not every category below is collected from every person. What we collect depends on the features you use, the permissions you grant, and whether you start a capture, memory, agent, custom-model, or developer workflow.

2. Information we collect

Account, access, and agreement information

We collect identifiers and profile information from sign-in providers, such as email address, phone number, display name, provider identifier, and authentication status. We also maintain waitlist and beta status, product entitlements, developer access, device and installation identifiers, agreement versions and hashes, acceptance timestamps, age confirmation, withdrawal and deletion status, and security records.

Messages, AI context, memories, and actions

When you use AI or agent features, we process the messages, prompts, files, images, transcriptions, responses, context, memories, preferences, corrections, instructions, and records of actions needed to provide the experience. If you ask Interfaces to remember something, we keep it until you forget it, delete it, or delete your account, subject to limited recovery and legal exceptions below.

Voice and Silent Speech data

Depending on the feature and device, we may process:

• microphone audio, speech transcripts, timing, language, and quality signals;
• front-camera video or frames showing your face and mouth;
• mouth-region crops, motion and timing features, model inputs, model outputs, and corrections;
• synchronized camera video, microphone audio, local transcript, timing, and quality data from completed Mac dictations; and
• prompted recordings, labels, derived features, evaluations, and personal-model artifacts if you choose to start custom-model training.

Face, mouth, and voice recordings may be sensitive personal information and may be treated as biometric identifiers or biometric information under some laws. Interfaces uses them to enable speech and interface research and product features, not to determine your identity, recognize you in public, or perform facial recognition.

The shared Silent Speech experience does not require a guided training set. Choosing not to start custom-model training stops that specific additional collection; it does not make research participation optional while you continue using the Interfaces research preview.

On iOS, some speech recognition may be performed by Apple under your device settings and Apple’s policies. Other voice features may route audio to Interfaces or an approved speech provider. A capture starts only when you invoke the relevant feature and grant the device permission it requires.

Keyboard data

The Interfaces keyboard handles ordinary key taps and local suggestions on device. Using its microphone or mouth control hands the requested capture to the Interfaces app. We do not collect everything you type merely because the keyboard is enabled. If you explicitly send text or context to an AI feature, that submitted content is handled as message and AI-context data.

Developer-platform data

We collect developer account and access information, API keys in hashed or protected form, projects and internal resource ownership, request metadata, usage and billing records, webhook configuration and delivery records, and the media or content a developer intentionally submits. Internal organization and project records may be used to enforce isolation even when they are not shown as separate concepts in the product UI.

Device, diagnostics, and analytics data

We collect limited device and app information, feature events, request identifiers, performance and latency measurements, error classifications, crash reports, coarse region, browser and operating-system information, and security signals. Product analytics are designed not to contain messages, transcripts, prompts, recordings, memory text, or API payloads. Our public website may also use cookies or similar technologies for analytics and essential operation.

Communications and payment information

We collect information you send to support, privacy, security, sales, or research contacts. If paid services are offered, payment processors handle payment-card details and provide us transaction, subscription, and billing status.

3. How we use information

We use information to:

• provide, personalize, secure, debug, and support the Service;
• authenticate accounts and enforce beta, developer, feature, and rate-limit access;
• transcribe speech, decode mouth movement, generate AI responses, maintain memories, and carry out requested actions;
• create and operate a custom model when you choose that feature;
• conduct research and develop, train, evaluate, test, and improve current and future speech, interface, agent, and AI models under the Research Agreement;
• understand reliability and product use without placing user content in analytics events;
• prevent fraud, abuse, unauthorized access, and safety or security incidents;
• communicate about the Service and respond to requests; and
• comply with law and enforce our agreements.

Our legal bases, where required, include performing our contract, your consent—including the Research Agreement and any device permission—our legitimate interests in operating and securing the Service, and compliance with law.

4. Research and model improvement

The entire account-based Interfaces platform is currently a research preview. The Research Agreement describes the information and purposes covered by the required research consent. We may use covered messages, responses, memories, actions, audio, video, transcripts, corrections, labels, derived features, model artifacts, and quality measurements to operate, personalize, research, train, evaluate, and improve Interfaces.

Not every feature collects every data type. For example, we do not collect a guided custom-model dataset unless you start that workflow. But there is not a second, optional “research contribution” consent for ordinary platform use.

We do not sell, lease, or trade face or voice recordings, or information derived from them, as data. We do not license them to data brokers or advertisers. We may charge for products and services that use models developed through the research. We may publish or share aggregate research results that do not reasonably identify you.

5. How we disclose information

We disclose information only as reasonably needed to the following categories:

• Cloud and infrastructure providers, including Google Cloud and Firebase, for storage, compute, authentication, delivery, monitoring, and security.
• AI and speech providers, such as Google, OpenAI, Deepgram, and ElevenLabs, when a selected feature is configured to use them. The provider and route may vary by modality, availability, quality, and account configuration.
• Platform providers, including Apple and Google, for app distribution, device permissions, speech services, notifications, authentication, and integrity checks.
• Analytics and reliability providers, including Google Analytics and Firebase, for content-free product analytics, crash reporting, and service health.
• Payment, communications, and professional providers that help us operate, support, bill for, or protect the Service.
• Authorities or other parties when required by law or reasonably necessary to protect rights, safety, security, and the integrity of the Service.
• A successor or transaction party in a financing, merger, acquisition, reorganization, or sale, subject to appropriate protections and notice where required.

We do not sell personal information for money and do not use personal data for cross-context behavioral advertising. We honor legally required opt-out preference signals. If applicable law treats limited analytics disclosure as a “sale” or “sharing,” you may opt out by contacting us as described below.

6. Retention and destruction schedule

We keep information only as long as reasonably needed for the purpose collected, then delete or de-identify it, subject to the following schedule and legal exceptions:

• Account and agreement records: while the account is active. Account deletion disables access promptly and ordinarily completes deletion of associated service data within 30 days. We may retain a minimal agreement, withdrawal, deletion, fraud, or security record when needed to demonstrate compliance and prevent reuse or abuse.
• Messages, memories, and agent records: until you delete or forget them, delete your account, or they are no longer needed for the experience. Encrypted recovery copies may remain for up to 7 additional days.
• Transient inference media: ordinary media used only to complete a synchronous request is not intentionally retained after processing. Temporary job, upload, retry, and quarantine objects expire automatically after the applicable workflow or short operational recovery period.
• Research recordings and derived artifacts: while your account is active and the information remains needed for the agreed research and product purposes, until you withdraw or request deletion, or until the purpose has been satisfied—whichever comes first. In all cases, biometric identifiers and biometric information are destroyed no later than three years after your last interaction with Interfaces, unless a shorter period applies.
• Custom-model data and artifacts: while you use the custom-model feature or until withdrawal, deletion, or the purpose is satisfied. Recovery copies may remain for up to 7 additional days.
• Mac capture queue: until upload succeeds, the capture is dismissed or deleted, the workflow reaches a terminal state, or local cleanup removes it.
• Developer request media and results: synchronous request bodies are streamed rather than intentionally archived; asynchronous source objects, results, and job records are removed under the endpoint’s documented lifecycle, account deletion, or developer deletion controls.
• Analytics, diagnostics, and operational logs: generally no more than 90 days, except that aggregated statistics may be kept longer when they no longer reasonably identify a person, and security-incident evidence may be kept as legally or operationally necessary.
• Support, billing, and legal records: for the period needed to resolve the request, meet tax and accounting obligations, enforce agreements, and comply with law.

When deletion begins, encrypted backups, soft-deletion systems, and disaster-recovery copies may persist for a short additional period before automatic removal. Data a user independently sent to or made public through a third party may remain under that third party’s controls.

7. Withdrawal and deletion

You may stop using the Service, withdraw from future research use, forget individual memories, or request account deletion through available product controls or by contacting privacy@interfaces.inc.

Withdrawal stops new research collection and future use of identifiable source data where technically and legally feasible. It cannot undo analysis or training already completed or remove your influence from a model version already trained and released. We will exclude deleted or withdrawn source data from future training runs and retain only the minimal pseudonymous record needed to honor the withdrawal and prevent future reuse.

8. Your privacy rights

Depending on where you live, you may have rights to know, access, correct, delete, or obtain a portable copy of personal information; withdraw consent; limit certain uses of sensitive information; opt out of sale, sharing, or targeted advertising; and appeal a decision. You will not be discriminated against for exercising a right.

Use in-product controls where available or email privacy@interfaces.inc. We may need to verify your identity and authority. An authorized agent may submit a request where permitted by law.

9. Security

We use administrative, technical, and physical safeguards designed for the sensitivity of the information, including encryption in transit, encryption at rest for cloud storage, access controls, scoped service identities, protected secrets, and monitoring. No system is completely secure, and we cannot guarantee absolute security.

10. International processing

Interfaces operates from the United States. We and our providers may process information in the United States and other countries. Where required, we use appropriate transfer mechanisms and safeguards.

11. Children

The account-based Service and research preview are for adults 18 and older. We do not knowingly permit a child to create an Interfaces account or submit research recordings. If you believe a child’s information was submitted, contact us so we can investigate and delete it.

12. Changes

We may update this Privacy Policy. We will provide notice of material changes through the Service, by email, or on this page. If a change requires renewed consent, you must accept the new version before continuing to use the relevant Service.

13. Contact

Dafdef, inc. d/b/a Interfaces
privacy@interfaces.inc