Privacy Policy

Areta AI Inc. (“Areta,” “we,” “us”) builds a wearable that listens and looks so you don’t have to remember everything yourself. That only works if you trust what happens to what it captures. This Privacy Policy explains what data the Areta wearable (“Device”), the Areta app (“App”), the Mori memory layer, our MCP server and APIs, our websites, and related services (together, the “Services”) collect, how we use it, who can see it, and the choices you have.

This Policy applies to users in the United States. Capitalized terms not defined here have the meaning given in our Terms of Service. If this Policy is translated, the English version controls.

Six things to know first

The Device records ambient audio and takes periodic low-resolution photos. That is the product. It captures your voice, the voices of people around you, and a rough picture of your surroundings.
Your voiceprint is what makes it yours. The Device needs to know which voice is yours. You enroll it once, with a separate consent, and can delete it at any time — but without it, own-voice mode and “who said what” labels stop working.
The light on the Device tells everyone what mode it is in. Light on = recording the whole scene. Light off = keeping only your own voice. The light adjusts its brightness to the environment so it is never intrusive, but it always stays visible and cannot be dimmed or turned off by you.
Anything that could identify a specific person is masked on your phone before upload. Faces, license plates, and similar identifying details are blurred on your iPhone. The camera is intentionally low-resolution and not designed to read text or documents.
Raw recordings are temporary; your Memories are what we keep. Audio and images are deleted once processed. What persists is the derived memory, which you can search, correct, export, and delete. Your data is encrypted in transit and at rest, but we do not offer end-to-end encryption: our servers must read your data to run AI on it.
We do not sell your data, do not show ads, and do not use your Memories to train models for other users. If that ever changes, it will be opt-in and we will ask you first.

1. Data we collect

1.1 Sensors on the Device

Sensor	What it collects	When it is active	What happens to it
Microphones	Ambient audio, including your voice and others’ voices	While the Device is worn and capture is on. Capture stops automatically after a period of inactivity, when the Device is taken off, or when you pause it.	Compressed on the Device, sent to your phone over encrypted Bluetooth, uploaded to our servers for transcription, then deleted (Section 8)
Camera (low-resolution)	Periodic still images of your surroundings. Not video. Not designed to resolve faces, text, or screens.	At fixed intervals while capture is on and the Device is not covered	Masked on your phone (Section 5), uploaded for scene understanding, then deleted
Motion sensor (IMU)	Whether the Device is worn, moving, stationary, or covered	Always while powered	Used on the Device to decide when to capture and to save battery; only summary state is uploaded
Device telemetry	Battery, firmware version, connection and capture state, error logs	Always while powered	Used to operate the Device and diagnose problems

The Device has no GPS, does not record video, and never streams audio live to anyone. The App does not collect immutable hardware identifiers of your phone (such as IMEI, IMSI, or SIM serial).

1.2 Recording modes and the indicator light

Full-scene mode (light on): everything the microphones hear is transcribed.
Own-voice mode (light off): the Device still listens, but only speech matching your voiceprint is kept; other voices are discarded before transcription.
If the Device detects that it is covered by clothing, it switches to own-voice mode automatically.
The light adjusts its brightness to ambient light so it is not distracting in a dark room, but it remains visible to others at all times. There is no setting to dim or disable it.

1.3 Data derived by our Services

Category	What it is
Transcripts	Text of what was said, with timestamps, produced by speech-recognition models we run ourselves
Speaker labels	Which speech is yours, based on your voiceprint (Section 4); other speakers are grouped within a recording and labeled by name only when you add a name
Memories	Summaries, memory cards, extracted tasks, people and places you mention, topics, and insights generated by AI
Long-term profile	A structured model of your goals, preferences, relationships, and recurring themes, built from your Memories over time and used to personalize answers and suggestions
Scene context	Approximate scene type inferred from masked images (e.g., office, restaurant, outdoors)
Text-based mood cues	Short status descriptions inferred from the words in a transcript (for example, noting that a conversation sounded tense). This is language analysis of what was said; it does not analyze your voice tone, face, or body. It can be turned off (Section 10.1).

1.4 Data you provide

Account details: name, email, password (hashed), profile photo if you add one
Voice enrollment sample used to create your voiceprint
Names you assign to other speakers; corrections to transcripts and Memories; notes, tasks, bookmarks
Questions you type or speak to the assistant, and photos you take or import for a question (chat records)
Photos from your phone’s library, only if you choose to import them (off by default)
Friends you add and content you choose to share
Purchase, shipping, and support communications
Third-party accounts you connect (Section 7), such as calendar or email. This includes account information and related content you authorize Areta to access, such as emails, calendar events, and files, depending on what you authorize and which features you use.

1.5 App permissions and what each is for

Permission	Why we ask	Default
Bluetooth	To connect to your Device	Required
Microphone	To ask the assistant a question by voice in the App	Asked when you first use voice chat
Camera	To take a photo for a question in the App	Asked when you first use photo chat
Location	To attach approximate location to Memories so you can search by place	Off
Photo library	To import photos into your Memories	Off
Notifications	Daily reviews, task reminders, Device alerts	Asked at setup

We ask for a permission only when you first use the feature that needs it. If you decline, we will not ask again for at least 48 hours, and declining never blocks unrelated features.

1.6 Collected automatically from the App

Phone model, OS version, App version, language, time zone; Bluetooth pairing identifiers for your Device
Usage analytics and crash reports (which features you use, errors, performance) that never include transcript or Memory content, and that start only after you have consented at sign-up
Payment data is processed by Apple or our payment processor; we receive purchase confirmation, not your full card number

1.7 Data about other people

In full-scene mode our Services process the voices and words of people around you, and masked images of the places where they are. We treat this data as part of your Memories. We do not build profiles of non-users, do not create voiceprints of non-users, do not attempt to identify non-users by voice or face, and do not contact them. You are responsible for recording others lawfully (Terms of Service, Section 5). See Section 11 for information addressed to people who are recorded.

1.8 Offline use

If your phone is out of range, the Device keeps capturing according to its current mode and stores encrypted audio in a small buffer. When the Device reconnects, the buffer is uploaded and processed, then deleted. If the buffer fills, older segments are overwritten. If you are not signed in to an account, the Device does not capture.

2. Consent and defaults

Nothing is collected before you agree. At sign-up you accept this Policy with a separate, un-prechecked checkbox. No data — including analytics or crash reporting — is collected until then.
Voiceprint consent is separate. Before voice enrollment, a dedicated screen explains what a voiceprint is, why it is required, how long we keep it, and how to delete it. You consent to it separately from this Policy (Section 4).
Non-core features are off by default and each has its own switch: full-scene recording, location tagging, photo-library import, integrations, social features, text-based mood cues, personalization.
Withdrawing consent is as easy as giving it — every switch is within three taps in Settings. Withdrawing consent to a non-core feature does not affect the rest of the Services. Withdrawing consent to the voiceprint disables own-voice mode and speaker attribution; other features keep working. We tell you this before you confirm.
If we materially change this Policy, we will show it to you in the App and ask you to accept again before continuing.

3. How we use data, including AI

Purpose	Data used
Operate the core product: transcribe audio, generate Memories, answer your questions, extract tasks, produce daily reviews	Audio, images, transcripts, Memories, profile, chat records
Recognize your own voice so transcripts are attributed correctly and own-voice mode works	Voiceprint, audio
Personalize insights and suggestions and maintain context across conversations (can be turned off)	Memories, long-term profile
Look up public information when you ask a question that needs it	Your question, with identifiers removed, sent to a web-search provider
Connect to third-party tools you authorize	Memories you choose to expose
Social features you choose to use	Content you choose to share
Ship Devices, provide support and warranty service, and (once we introduce paid plans) process payments	Account, purchase, and support data
Keep the Services secure, prevent fraud and abuse, debug problems	Telemetry, usage analytics, logs
Improve the Services	See 3.2
Comply with law and enforce our Terms	As required

3.1 How the AI pipeline works

Every conversation you capture goes through the same steps. Everything the Services show you that came out of this pipeline is labeled as AI-generated.

Speech recognition turns audio into text. This runs on models we host ourselves; audio does not leave our infrastructure for this step.
Speaker separation groups the speech by voice and marks which parts match your voiceprint. Also self-hosted.
Scene understanding sends the transcript and masked images — with your account and device identifiers replaced by a temporary session token — to one of our large-language-model providers (Google Gemini, OpenAI, or Anthropic), which returns summaries, tasks, and memory cards.
Assistant answers your questions using your Memories and profile, through the same providers under the same protections.

Each user’s Memories are isolated by account and session; the assistant can never draw on another user’s data.

3.2 How we improve the product without reading your life

We use aggregated and de-identified usage data (for example, how often a feature is used, transcription error rates) to improve the Services.
Our staff do not routinely read your transcripts or Memories. Access to content is limited to what is needed to resolve a support request you initiate or to investigate a security incident, requires approval, and is logged for at least 12 months.
We do not use your audio, images, transcripts, or Memories to train AI models that serve other users. If we ever offer that, it will be a separate opt-in that is off by default, uses only de-identified data, and can be withdrawn at any time.
Data obtained through Google Workspace APIs, and content derived from it, is not used to develop, improve, or train generalized or non-personalized AI models, and is not eligible for the model-improvement opt-in described above.

3.3 What we do not do

We do not collect health data and do not make health inferences. The Services are not a medical device and are not HIPAA-compliant; do not use them to store or process protected health information (see Terms of Service, Section 1).
We do not analyze your voice tone or face to infer emotion. Text-based mood cues (Section 1.3) are optional and based only on the words spoken.
We do not make decisions about you with legal or similarly significant effects (such as credit, employment, housing, or insurance). Memories and insights are suggestions; you can correct or delete any of them.
We do not perform real-time remote identification of anyone.

4. Voiceprint and biometric data

To tell your voice apart from others, we create a voiceprint: a mathematical representation of the characteristics of your voice. This may be considered biometric information under some state laws, including the Illinois Biometric Information Privacy Act, the Texas Capture or Use of Biometric Identifier Act, Washington’s biometric law, and the California Consumer Privacy Act. This section is our written biometric data policy, including our retention schedule and destruction guidelines.

Why it is required. The voiceprint is how the Device knows which speech is yours. Without it, own-voice mode cannot work and transcripts cannot be attributed. It is a core function: you can use the Services without it, but transcripts will not distinguish your voice from others and own-voice mode will be unavailable.
Consent. We create your voiceprint only after you complete voice enrollment in the App and give explicit written consent on a dedicated screen, separate from your acceptance of this Policy. That screen states the purpose, retention period, and deletion method, and which features are unavailable if you decline.
Purpose. Your voiceprint is used solely to label which speech in your recordings is yours and to operate own-voice mode. It is not used to identify you elsewhere, to identify anyone else, or shared with third parties for any purpose. Voiceprint matching runs on models we host ourselves.
Storage and retention. Voiceprints are encrypted with a separate key and stored separately from your Memories. We permanently destroy your voiceprint when you delete it in Settings, when you delete your account, or within three (3) years of your last interaction with the Services, whichever comes first, unless a shorter period is required by law. Deleting your voiceprint disables own-voice mode and speaker attribution until you re-enroll; everything else keeps working.
No sale, lease, or trade. We never sell, lease, trade, or otherwise profit from biometric data.
Other people’s voices. We do not create or store voiceprints of other people. Speaker labels for others are based on grouping within a single recording and names you assign, not on stored biometric templates.

5. Images and masking

The Device camera is intentionally low-resolution and captures still images at intervals, not video.
Before any image leaves your phone, on-device software masks anything that could identify a specific person — faces, license plates, name badges, screens, and similar identifying details. Raw, unmasked images are never uploaded to our servers.
Masked images are used to add scene context to your Memories and to help the AI understand what happened, then deleted on the schedule in Section 8.
Photos you import from your phone library are not automatically masked, because you chose them; treat them as you would any photo you share.

6. Who we share data with

We do not sell personal information and do not share it for cross-context behavioral advertising. We share data only as follows. Every provider that processes personal data on our behalf is bound by a written data-processing agreement that limits it to our instructions and requires deletion on request.

Recipient	What	Why
Cloud hosting and infrastructure (Amazon Web Services)	Encrypted data at rest and in transit	To run the Services
Large-language-model providers (Google Gemini, OpenAI, Anthropic)	Transcripts, masked images, and questions, with your identifiers replaced by a session token. When you use connected services, also the related content needed to answer your question or complete an action you request.	Scene understanding and assistant answers, under contracts that prohibit each provider from using your data to train its models or retaining it beyond the request
Content search and data processing providers	Content from the emails and documents you authorize access to, as needed to provide search	To help you search and find content in your connected services
Web-search provider (Brave Search API)	Search queries derived from your question, without identifiers	To answer questions that need public information
Apple and Google	Device identifiers and pairing metadata	Bluetooth accessory pairing, App Store distribution, and device-finding features you enable
Payment processors and shipping carriers	Purchase and delivery data	To sell and ship Devices and, later, to bill paid plans
Analytics and crash-reporting providers	Usage and diagnostic data (never transcript or Memory content)	To keep the App reliable
Integrations you authorize	The Memories you choose to expose via MCP, API, or export	Only at your direction; governed by that third party’s policies once transferred
Friends you share with	Content you choose to share	Only at your direction
Professional advisors	What is necessary	Auditors, lawyers, accountants, under confidentiality
Successor entities	All categories	In a merger, acquisition, financing, or sale of assets; we will notify you and honor this Policy or give you a chance to delete your data. Google user data is transferred only as permitted by the applicable Google data policies, and any transfer as part of a merger, acquisition, or sale of assets requires your prior explicit consent.
Law enforcement and legal process	What is legally required	We require valid legal process, notify you when permitted, and publish a transparency report

Speech recognition and voiceprint matching run on our own infrastructure and are not shared with any provider. A current list of third-party SDKs in the App is available in the App’s Third-Party Notices.

7. Integrations, MCP, and exports

You can connect your Memories to third-party AI tools (for example, through our MCP server) and export them in standard formats. This is a core part of the product: your memory should work in the tools you already use.

Integrations are off by default. You choose what to expose, per integration, and can revoke access at any time in Settings.
After data is transferred to a third party at your direction, that party’s privacy policy governs it. The restrictions on the use and transfer of Google user data and content derived from it continue to apply when that data is processed by a third party. Choose your tools carefully.

7.1 Third-party integrations

When you connect a third-party service, Areta accesses related content within the permissions you grant and uses it to search, generate responses, and complete actions you request. To complete an action, Areta may send the necessary content and instructions to that service, for example to send an email or change a calendar event. Some connected features keep syncing related content while access is authorized, until you disconnect the service or revoke access.

If you connect Google services (such as Calendar or Gmail), our use and transfer of data received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements.

7.2 MCP

When a third-party tool queries your Memories through MCP, we record call metadata for usage statistics and troubleshooting. This includes account and connection identifiers, call time, tool name, status, duration, argument names, time filters, and result metadata such as counts and memory identifiers. These call records do not contain search text, Memory content, or transcript content. You can view usage statistics in Settings.

7.3 Exports

You can export your Memories in standard formats and keep them yourself or use them with other tools. Keep exported files secure. If you give a file to a third party, that party’s privacy policy governs how it handles the file.

Revoking an integration’s access does not recall files you have already exported and does not delete copies kept by you or by third parties.

8. Retention and deletion

Data	Retention	Your controls
Emails, calendar events, and documents from connected services, and related data used for search	Kept to provide the sync, search, and assistant features you use. Retention varies by data type and by whether the service is still connected.	Disconnect the service on the Plugins page in the App to stop further syncing. Disconnecting does not delete all stored data.
Raw audio	Deleted from our servers as soon as transcription completes; deleted from your phone after upload	Delete a conversation and its audio at any time
Masked images	Deleted as soon as scene understanding completes	—
Transcripts and Memories	Kept until you delete them or your account	Delete a single Memory, a day, a date range, or everything
Long-term profile	Kept until you delete your account	View, edit, or reset in Settings
Voiceprint	Until deleted or account deletion; maximum 3 years inactive	Delete in Settings (disables own-voice mode and speaker labels)
Chat records with the assistant	Kept until you delete them or your account	Delete per conversation
Location tags	Attached to Memories; deleted with them	Turn off in Settings
Account and purchase records	Life of account plus as required by tax and accounting law (typically 7 years)	—
Support communications	3 years	—
Access audit logs	12 months minimum	—
Backups	Deleted data is removed from backups within 35 days	—

Automatic expiry. Data past its retention period is deleted or anonymized by scheduled jobs; nothing depends on someone remembering to do it.
Deleting your account removes your Memories, profile, voiceprint, chat records, and raw captures from the App, our databases, caches, and logs immediately, and from backups within the period above. We also instruct our processors to delete your data and record their confirmation. We may retain limited records (for example, purchase history, the fact that a deletion request was made) as required by law.
Corrections propagate. When you correct a transcript or delete a Memory, derived summaries and your profile are updated so the error does not persist downstream.
Factory reset and resale. Resetting a Device wipes its local buffer, unlinks it from your account, and tells our servers to stop accepting uploads from it. A Device must be unlinked before it can be paired to another account, so a previous owner’s data is never visible to the next.
Lost Device. Unlinking prevents further uploads. Audio buffered on a Device that has not synced is encrypted and cannot be read without the paired phone.
Planned: no-cloud-retention mode. We are building a mode in which our servers process each conversation in real time and return the result to your phone without storing any Memory in the cloud. We will update this Policy when it ships.

Revoking access from within a third-party account does not automatically delete content Areta has already stored. When you disconnect a service in Areta, we delete that connection’s authorization data and the synced copies of emails, calendar events, and documents. Existing chat records and content saved separately or exported are not automatically deleted, and the original content in the third-party account is not changed.

9. Security

Credentials for third-party services are stored encrypted and bound to the connection they belong to.
All data is encrypted in transit (TLS 1.2 or higher, no fallback to unencrypted connections) and at rest (AES-256). Device-to-phone transfer uses Bluetooth LE Secure Connections and Apple’s accessory authentication.
Each user’s data is encrypted with per-user keys and isolated by account, so that deletion is verifiable and one account’s data cannot be read through another’s. Voiceprints use a separate key.
Access to production data is role-based, limited to the minimum necessary, logged, and reviewed. Backups are encrypted and stored separately from production; we test restoring them regularly.
The Device receives firmware updates over the air. Security fixes for critical vulnerabilities are installed automatically; you cannot defer them. We scan our software dependencies continuously for known vulnerabilities.
We maintain a written incident-response plan. If a breach affects your data we will notify you and regulators as required by law, and no later than 72 hours after we confirm it where the law so requires.
We are pursuing SOC 2 certification and will publish our status at areta.ai/trust.
What we do not offer: end-to-end encryption. Our servers must decrypt your data to run AI on it. If you need a product where no one but you can ever read your recordings, Areta is not that product today.

10. Your rights and choices

10.1 Controls in the App (each within three taps of Settings)

Pause capture on the Device or in the App; switch recording mode between full-scene and own-voice (the light reflects the mode).
Review, correct, and delete any transcript, Memory, chat record, or profile entry.
Turn off personalization: the assistant stops using your long-term profile and scene tags to shape answers and suggestions.
Turn off text-based mood cues.
Location tagging, photo-library import, integrations, social features, and any future model-improvement opt-in are each controlled separately.
See what we hold and export it at any time in standard formats.
Delete your voiceprint or your account.
Push notifications and marketing email can be turned off in phone settings or via the unsubscribe link.

10.2 Rights under U.S. state privacy laws

Depending on where you live (including California, Colorado, Connecticut, Delaware, Oregon, Texas, Utah, Virginia, and other states with comprehensive privacy laws), you may have the right to:

Know / access the personal information we hold about you and how it is used and shared;
Delete your personal information;
Correct inaccurate personal information;
Port your data in a usable format;
Opt out of the sale of personal information, sharing for targeted advertising, and profiling with legal or similarly significant effects (we do none of these);
Limit the use of sensitive personal information to what is necessary to provide the Services;
Not be discriminated against for exercising these rights.

To exercise these rights, use the tools in Settings or email support@areta.ai. Our process is: receive → verify identity (by confirming access to your account email) → act → reply. To disconnect a connected service, use the Plugins page in the App; Section 8 describes what data is deleted. You may designate an authorized agent by providing written authorization. We respond within 45 days and may extend once by 45 days with notice. If we deny a request, you may appeal by replying to our response; we answer appeals within 45 days, and if the appeal is denied you may contact your state attorney general.

10.3 California residents — additional disclosures

Categories collected (past 12 months): identifiers; commercial information; internet or device activity; geolocation (if enabled); audio and visual information; biometric information (voiceprint); inferences drawn from the above; and sensitive personal information (contents of your recordings, biometric data, precise geolocation if enabled, account credentials).
Sources: you, your Device, your phone, and service providers.
Disclosed for business purposes to: the service-provider categories in Section 6.
Sale / sharing: we do not sell personal information and have not done so in the past 12 months. We do not share personal information for cross-context behavioral advertising. We do not knowingly sell or share the personal information of anyone under 16. A “Do Not Sell or Share My Personal Information” link is provided on our website for completeness.
Sensitive personal information is used only to provide the Services you request, to maintain security, and as otherwise permitted by CPRA §7027(m). Because processing recordings is the Service itself, limiting that use means turning off capture, which you can do at any time.
Retention: Section 8. Shine the Light: we do not disclose personal information to third parties for their own direct marketing.
Global Privacy Control: we honor GPC signals as an opt-out of sale/sharing on our website, although we do not sell or share.
Financial incentives. If we offer referral rewards or discounts in exchange for personal information such as an email address, we will describe the terms at the time; participation is voluntary and you can withdraw at any time via the unsubscribe link or by contacting us.

10.4 Illinois, Texas, and Washington residents

Section 4 serves as our written biometric data policy, including our retention schedule and destruction guidelines, and is publicly available on our website. Illinois residents: the dedicated voiceprint consent screen provides the written notice and written release required by BIPA §15(b).

11. If you are near someone wearing Areta

If someone you are talking with is wearing an Areta Device, here is what happens to your data:

Look at the light. If it is on, the Device is transcribing the whole conversation, including you. If it is off, only the wearer’s own voice is kept and yours is discarded before transcription. The light adjusts to ambient brightness but is always visible; wearers cannot dim or disable it.
In full-scene mode your voice and words may be recorded and transcribed as part of the wearer’s Memories. Images of the scene are captured at low resolution, and your face and other identifying details are masked on the wearer’s phone before anything reaches us.
We do not create a voiceprint or profile of you, do not attempt to identify you, and do not contact you. Raw audio is deleted once transcribed (Section 8); the transcript belongs to the wearer.
You can ask the wearer to switch to own-voice mode or pause, or to delete the conversation. The wearer is responsible for complying with recording-consent laws.
If you believe you were recorded unlawfully, or wish to request removal of content about you, contact support@areta.ai. We will review the request and, where we can identify the content, work with the wearer or act as the law requires.

12. Children

The Services are not directed to, and we do not knowingly collect personal information from, anyone under 18; the App is rated 17+ in app stores. If you believe a child has created an account, contact us and we will delete it. Because the Device records ambient audio in full-scene mode, it may capture the voices of children around the wearer; that data is part of the wearer’s Memories and subject to the wearer’s controls and responsibilities.

13. Website, cookies, and marketing

Our website uses essential cookies and privacy-respecting analytics. We do not use advertising cookies or pixels that track you across other sites. If that changes, we will update this Policy and provide an opt-out.
If you subscribe to updates or buy a Device, we may send product and marketing email; unsubscribe at any time via the link in each email. Transactional email (receipts, security notices) is sent regardless.
We honor Global Privacy Control. We do not respond to “Do Not Track” signals because there is nothing to opt out of.
The App does not track you across other companies’ apps or websites, so it does not request App Tracking Transparency permission.

14. Where your data is stored

Areta AI Inc. is a U.S. company. Your data is stored and processed on Amazon Web Services infrastructure located in the United States. Our AI providers also process data in the United States. Our affiliate in Singapore may access data for operations; wherever data is processed, it is protected by the contractual and technical safeguards described in this Policy. Memory content is not transferred to our engineering contractors in mainland China.

15. Changes to this Policy

We will post updates here and update the “Last updated” date. For material changes — especially any change to how we use recordings, images, or biometric data, to where data is stored, or to our commitment not to train models on your Memories — we will notify you by email and in the App at least 30 days before the change takes effect and ask you to accept the updated Policy before continuing to use the Services.

16. Contact us

Areta AI Inc. — Attn: Privacy Lead
8 The Green, Ste A
Dover, Kent County, Delaware 19901
United States

Privacy requests and support: support@areta.ai
Privacy Lead and legal: legal@areta.ai