My Voice Agents

Your information

Privacy policy

Effective · iPhone and Android · Ages 18+

This policy explains what My Voice Agents processes, why it does so, who receives information, and how you can manage it.

Your workspace is stored on your device. Live AI sends context to OpenAI. Voice calls also send microphone audio directly to OpenAI. Deleting local content does not automatically erase information already processed by a provider.

1. Who we are and scope

Neu Software LLC, a limited liability company registered in Delaware, United States, operates My Voice Agents. “We,” “us” and “our” mean Neu Software LLC. Where applicable data protection law uses that term, we are the controller of personal information processed for the purposes we determine under this policy.

This policy covers our mobile app, myvoiceagents.app, our API at api.myvoiceagents.app, and support communications. For privacy questions or requests, email tash@neu.ie.

The app is for adults and is distributed for iPhone through the Apple App Store (including Apple TestFlight before public release) and for Android through Google Play. There are no user accounts: each installation connects automatically to our service using a random credential. Cloud workspace accounts and cross-device synchronization are not offered. The optional Pro subscription is bought through the App Store or Google Play, and each purchase requires your confirmation in the store.

We manage the API connection and model selection. The app does not ask you to enter a server URL, connection token or OpenAI API key. Provider credentials remain on our server.

2. Information and its sources

Information is supplied by you, selected from your device or a file provider, generated by the AI features you use, or produced by devices and networks when they connect. A document, shared agent or conversation you provide can also contain information about another person.

We do not require a personal name or email address to create a local agent. We do not purchase consumer profiles or import your address book. If you include sensitive or third-party information in content, it can follow the same processing described here; see section 15.

3. Your device and service access

The app stores agents, messages, saved memories, file references, extracted text and settings in its application storage. It keeps a current workspace and recovery snapshots to handle interrupted saves. Selected files are copied into an app-managed directory so the agent can use them later. Picker copies and exported files can also exist in device cache.

A random installation credential is stored separately using the operating system’s secure storage. Our server stores its hash, a daily network hash used to limit registrations, and usage records such as the current plan, chat-message and text-operation counts, call identifiers, timestamps and reserved or used seconds. If you subscribe, the server also stores the verified plan and expiry, billing cycle, a hash of the original store order reference or Apple original transaction identifier, and associations between installations restoring the same purchase. These records support access and usage limits; they do not contain a cloud copy of your conversations or files.

Text conversations and voice calls use OpenAI. The app does not use device text-to-speech to generate agent replies. Voice previews receive OpenAI audio without using the microphone. Browsing and editing local agents does not itself send their content to OpenAI.

Existing messages, memories and files can become context when you send a message or start a call. Opening external links and sharing exports can also send information to the services you choose.

4. Live AI, files and voice

AI messages and service processing

In iOS version 1.0.6 build 6 and later, the app shows an Enable AI conversations disclosure before the first AI message, voice call or voice preview. It names OpenAI, explains the content and technical information shared, and asks you to choose Allow sharing with OpenAI or Not now. No AI content is sent before you allow it. You can decline and continue browsing and editing your local workspace. Your permission is saved on this device for future AI use and can be withdrawn in Settings → AI sharing. Older builds may not include this control; update to a supported build or stop using live AI to stop sending new content. Microphone permission is requested separately before a call and does not replace this AI-sharing choice.

For a live chat, the app sends the agent's instructions, a bounded recent conversation, relevant saved memories when enabled, and selected file context to our My Voice Agents API. Our API sends the necessary content to OpenAI to produce a response and, when memory is enabled, possible memory suggestions. Replies return to the app and are stored with the local conversation.

Web search and page reading

Agents can search the public web and read public pages during chat or a call. Our server runs these tools using the same SearXNG service used by User Evaluation and a server-side page reader. Search queries are sent to the search engines configured in SearXNG. Requested URLs are fetched from the websites concerned, which receive our server's network information. Public search snippets and page excerpts are passed to OpenAI to help answer your request; source links may be saved with your local conversation.

We instruct agents to keep queries focused on the public topic and exclude credentials, private files, saved memories and unnecessary personal details. Model-generated queries can still contain information from your request. Do not ask agents to search for sensitive information you do not want shared with search services. Opening a source link connects your device to that website under its own privacy policy.

Attached files

Attaching a file initially creates a local copy. Sending a live message or starting a live call can send its contents, not just its filename. Chat selects a limited set of attached files using the current message’s words and recency; a new voice call selects up to five of the most recently added files; you do not receive a new file-picker prompt each time those files are reused. Text excerpts, images and PDFs may be included. An image or PDF for a voice call can first be sent through OpenAI's text-and-image processing to prepare a summary, which is then included in the voice session. Such summaries can omit details or contain errors.

Live calls and transcripts

Our API prepares the session context and creates an OpenAI Realtime call using credentials held on our server. The app receives connection-negotiation data, not our OpenAI API key. Microphone audio travels directly from the device to OpenAI over WebRTC; OpenAI returns audio and text. Our server meters call time and ends calls when their allowance or connection lease expires. Muting stops new microphone audio, and ending a call closes its connection.

The app processes transcripts during a call even if Save call transcripts is off. That setting controls whether call text is added to your persistent local chat history, including messages typed during the call. It does not disable OpenAI's processing, transcription or applicable retention. A local call-ended entry, date and duration can still be saved.

If agent memory is enabled, the app can send a temporary transcript through our API to OpenAI when a call ends to extract memories for automatic saving, including when transcript saving is off. The app and our API do not intentionally create a permanent raw-audio recording. This does not mean OpenAI retains no audio or other content under its own controls.

5. Memory and personalization

When agent memory is enabled, the app automatically saves eligible details after a completed chat reply or after a voice call ends, without a separate approval prompt. You can also add entries yourself. The app lets you view, edit and delete entries. Our model instructions and filtering try to keep suggestions grounded in your own words and avoid sensitive information; these measures cannot guarantee that every suggestion is accurate or appropriate.

With memory enabled, saved entries help tailor future responses and may be sent with relevant context. Turning memory off stops inclusion of saved memory entries in future requests and pauses automatic saving. It does not delete those entries or remove the same facts from instructions, past messages, files or a session that has already received them.

The AI may summarize or draw inferences from content to answer you, and those inferences may be wrong. We do not use this personalization to decide eligibility for employment, credit, housing, insurance or other opportunities with legal or similarly significant effects. The preview does not offer voiceprint authentication or identity matching; the voice feature is used for conversation and transcription.

6. Sharing and exports

Sharing an agent link, JSON definition or HTML website preview includes the agent's name, role, instructions, selected voice and appearance. It excludes attached files, saved memory entries and conversation history. Personal information placed in the definition itself still travels with it. A recipient can read that definition, including instructions encoded in an agent link, so review it before sharing.

Exporting a transcript includes the conversation text and associated timestamps. Sharing an individual file provides its contents to the application or recipient you select through the device's sharing tools. Those recipients can retain or forward their copies, and deleting the source in My Voice Agents does not recall them.

An exported HTML page is a preview, not a deployed live AI service. Its sample conversation does not call our API. If you publish it using another host, that host and any changes you or others make to the page determine the website's additional processing. There is no public agent directory or automatic publication of your local workspace in this preview.

7. Website, cookies and advertising

Cloudflare serves our website and its assets and provides DNS. Website delivery and security involve technical information such as IP addresses, requested URLs, browser or connection information, timestamps and error or traffic metadata. Cloudflare may make operational traffic information available to us through its service. Its network-error reporting can receive browser reports about failed connections.

Our current website code has no submission form or advertising tags. It counts page views with Umami, a cookieless analytics tool we host ourselves, which records the page, referrer, browser, device type and country without identifying you. It does not set advertising or analytics cookies or use browser storage for those purposes. The interactive examples stay in the page and do not send your example text to an AI model. Browsers can cache assets, and hosting or security services can use technical mechanisms needed to provide their service; this policy does not promise that every browser or provider is cookie-free.

The app contains no advertising, analytics or crash-reporting SDKs; Umami runs only on this website. We do not sell personal information, share it for cross-context behavioral advertising, or use it for targeted advertising. Browser signals such as Global Privacy Control do not change these current practices because those activities are not performed. Legacy “Do Not Track” signals likewise do not change the operation of the app or website.

8. Recipients and other disclosures

We require third parties receiving personal information from us to provide the same or an equivalent level of protection as this Privacy Policy and the applicable App Store privacy requirements, including appropriate security, confidentiality and limits on use. Provider-specific processing and retention are described below; a provider’s separate policy does not remove our responsibility for the information we share.

OpenAI

OpenAI receives the content and technical information needed for live AI, including the request context, live voice and transcripts described above. It processes that information to deliver the models and operate and protect its services. Our text, memory-extraction and file-summary requests use store: false.

OpenAI states that API content is not used to train its models by default unless the API customer opts in. Its default abuse-monitoring retention is up to 30 days and can include content; legal or safety exceptions can extend this. Some prompt caching can retain encrypted representations for up to 24 hours, and flagged image or file inputs can be retained for safety review. store: false does not remove all provider retention, and we do not represent the Service as having Zero Data Retention. See OpenAI's API data controls.

Infrastructure providers

Hetzner hosts our API infrastructure. The API receives live request content to perform the requested work; it does not maintain server-side conversation history or a user file library. Network and infrastructure providers necessarily handle traffic and operational information. See Hetzner's privacy information.

Cloudflare delivers our website and DNS, as described above. The API hostname currently uses DNS-only routing: its HTTPS traffic goes directly to our API host, not through Cloudflare's reverse proxy. That distinction does not describe OpenAI's own provider infrastructure. See Cloudflare's privacy policy.

Device services, support and other recipients

Your operating system, file provider, backup service, browser, sharing applications and any distribution platform can process information when you use their features. Opening external links is governed by the destination service's practices. These parties do not receive your full workspace from us simply because you install the app.

If you contact us, our email service and people handling the request process the details you send. We may involve a relevant service provider to investigate an issue, using the information needed for that investigation. Avoid sending access tokens, model API keys, government identifiers, private documents or a complete workspace when a brief description will do.

Where reasonably necessary and legally permitted, information may be disclosed to professional advisers or appropriate authorities to comply with binding law or valid legal process, protect rights, investigate abuse, or address security incidents. If the service is involved in a business transfer, information under our control may be disclosed to advisers and an appropriate successor, subject to applicable notice and privacy obligations. This does not authorize unrelated advertising use.

Purchases

Pro subscriptions are sold only through the Apple App Store on iPhone and Google Play on Android. The store handles the purchase, payment method, receipts, renewals and refunds under its own terms and privacy policy. Agent instructions, conversations, files, memories and microphone audio are never part of billing.

Apple App Store: when you buy or restore on iPhone, the app sends our server the signed transaction information Apple provides for that purchase. Our server verifies Apple’s signature on it and checks it with Apple’s App Store Server API, including the transaction and original transaction identifiers, product, purchase and expiry dates, and whether it was refunded or revoked, and may receive Apple’s App Store Server Notifications about renewals, cancellations and refunds. We associate the purchase with your installation using a pseudonymous identifier derived from it. See Apple’s privacy policy.

Google Play: our server sends an installation-linked pseudonymous identifier with the purchase and verifies subscription purchase tokens directly with Google Play. We store encrypted purchase tokens. Google sends subscription notifications to our server through Google Cloud Pub/Sub. See Google’s privacy policy.

For both stores we store the product, subscription status, expiry and usage records to provide, restore and manage access, prevent duplicate allowances and handle refunds or disputes. Earlier test-only upgrades do not create a paid subscription. They do not authorize a later charge or automatically convert into a purchase.

9. Purposes and legal bases

We use the information described above to provide the functions you request, maintain the workspace, deliver the AI features you request, authenticate installation access, verify and restore subscriptions, keep the service reliable and secure, respond to support and privacy requests, and meet legal obligations. We do not use your conversations to build an advertising profile.

Where EEA or UK data protection law applies, the legal bases for our processing are:

You are not required by law to submit content to the app. Live features need the relevant content, network connection, permission and access credential to work; a voice call also needs microphone permission. A support or privacy request may need enough information for us to understand it and, where necessary, verify the requester. These bases do not mean that any sensitive information is suitable for the preview.

10. Retention and deletion

Your workspace and files

Local information remains until you remove it, reset the workspace or clear the app's device data. You can delete memories, remove files, delete agents or select Settings → Reset this device. A previous recovery snapshot may still contain removed text until a later save overwrites it or a full reset clears the snapshots.

Reset clears managed workspace snapshots and imported agent files. The installation credential and server-side usage records remain so resetting local content does not refill an allowance or cancel a subscription. Reset cannot erase exports, recipient copies, device backups or information already processed by providers. Picker and export cache files may remain until cache or app data is cleared. Contact us about deletion of information under our control; device or storage failures can prevent local deletion from completing.

Server processing and operational logs

Our private API handles request bodies, attached content, prepared summaries and replies while fulfilling a request, without intentionally persisting them as a conversation database. Request and response buffering to disk is disabled at its reverse proxy. Temporary in-memory processing is not a secure-erasure guarantee.

Routine API application logs contain request identifiers, methods, recognized route names, response status, duration and sanitized failure information. They are configured to exclude content, authorization credentials and query strings. Routine API access logs contain timestamp, method, status and duration, excluding full URLs and client IP addresses. IP addresses are still processed for connections and request limits; a daily network hash is retained with installation registration records. Rare critical web-server errors can produce additional connection or request diagnostics.

API web-server logs rotate daily when nonempty, retaining 14 rotated files plus the active log. Empty files are not rotated, so this is not a guaranteed 14-day maximum age. Container logs rotate across three files of up to 10 MB each; their maximum age depends on volume and is not time-limited by that setting.

Installation, usage and verified subscription records remain across app restarts and local resets to enforce allowances, restore access and prevent duplicate credits. They do not yet have an automatic expiry. Contact us to request deletion, subject to applicable security, usage-accounting and legal requirements.

Providers, support and legal records

OpenAI retention is described in section 8. Apple, Google Play and Google Cloud (for purchases and billing notifications), Cloudflare, Hetzner, device services and other recipients apply their relevant operational, security and legal retention practices. We do not promise one common retention period across them.

We keep support and rights-request correspondence for as long as reasonably needed to resolve the request, handle related follow-up, document our response, address disputes or satisfy applicable legal obligations. The criteria include the issue's status, sensitivity, continuing security needs and any applicable recordkeeping requirement. Information retained for a legal obligation or dispute is limited to that purpose.

Deleting a local conversation does not automatically submit a deletion request to OpenAI or other recipients. Contact us about information under our control or requests involving our providers. We will explain applicable exceptions and what we can identify; we cannot search or restore your local-only workspace from our server.

11. Your controls

Changes to stored content cannot retract data already supplied to a live session. End that session before changing context you want excluded from the next one. Contact us if you cannot use a control or need an accessible way to make a request.

12. Security

The hosted website and API use HTTPS, and live audio uses encrypted WebRTC transport. Model requests require an installation credential, and OpenAI credentials remain on our server. The app stores its installation credential in operating-system secure storage. Usage limits and sanitized logging reduce unnecessary exposure.

The API and AI provider must be able to process the content sent to them. This is not end-to-end encryption that prevents those services from accessing it. Ordinary local workspace files rely on device and application-storage protections; the app does not put them all in its secure credential store. No network, device or service can be guaranteed completely secure.

Protect your device and avoid sharing credentials in messages or support requests. If you suspect unauthorized access or a security incident, contact tash@neu.ie with a brief description. We will handle any applicable notification obligations under the law that applies.

13. International processing

Neu Software LLC is registered in the United States. Our infrastructure and AI providers operate internationally, and information can be processed in the United States and other countries outside your own. Those countries may have different privacy laws.

The location of an API server does not determine where all OpenAI, Cloudflare, support or device-provider processing occurs. We do not promise that information stays in a particular country or region. For information about relevant processing locations, the transfer arrangements applicable to your information, or access to any applicable transfer safeguards, contact us using the details below.

14. Privacy rights and requests

How to contact us

Email tash@neu.ie and describe the information and action involved. You can use the subject “Privacy request,” but no special wording is required. Include your country or state if useful for identifying applicable rights. Do not send a password, connection token or complete identity document with an initial request.

Rights vary with applicable law and our role. We may need proportionate verification before disclosing, changing or deleting information. We will request only what is reasonably needed and use verification information for that process. We will explain any refusal, limitation or permitted extension and respond within the period required by applicable law. Requests are normally free, subject to exceptions permitted by that law.

We cannot remotely access a local workspace that has not been sent to us. We can guide you through local controls and address information we or our providers can identify. Please do not send additional private content solely to establish that a local file exists.

EEA and United Kingdom

Where applicable, you may request access to personal data, correction, erasure, restriction and a portable copy. You may withdraw consent at any time. These rights have conditions and exceptions; for example, erasure can be limited by a legal obligation to retain a record.

Your right to object: you may object to processing based on legitimate interests for reasons relating to your circumstances. We will assess the objection as required by applicable law. You may object to direct marketing at any time; we do not currently use your data for that purpose.

You may complain to a competent data protection authority, including where you live or work or where an alleged infringement occurred. See the EEA supervisory-authority directory and the UK Information Commissioner's Office. Contacting us first is welcome but does not remove your complaint rights.

United States state privacy laws

Where a state law applies, rights may include confirming processing; accessing categories or specific pieces of personal information; correcting inaccuracies; deleting information; obtaining a portable copy; and learning about sources, purposes and categories of recipients. Some states also provide access to a list of specific third-party recipients or categories of such recipients.

Applicable laws may provide rights to opt out of sale, sharing for cross-context behavioral advertising, targeted advertising or profiling used for legal or similarly significant decisions, and to limit certain uses of sensitive information. We do not carry out those sale, advertising or significant-decision profiling activities. We use sensitive information supplied in content for the requested function and applicable operational or legal purposes, not advertising. You may still contact us to exercise any right that applies.

We will not unlawfully discriminate or retaliate against you for exercising privacy rights. A feature may nevertheless be unavailable where the information or permission you choose not to provide is necessary for that feature.

Authorized agents and appeals

Where applicable, an authorized agent can submit a request for you. We may ask for evidence of authority and verify your identity directly where the law permits, without imposing requirements on opt-out requests that the law prohibits.

If we decline a request and an appeal right applies, reply to our decision or email tash@neu.ie with “Privacy appeal,” identifying the decision and why you want it reviewed. We will review the appeal and provide a written explanation within the applicable deadline. If an appeal is denied, you can contact your state attorney general or other regulator, and we will provide the complaint information required by law. General information is available from the California Attorney General and in the Delaware Personal Data Privacy Act.

15. Children and sensitive information

My Voice Agents is intended for people aged 18 and over. It is not directed to children, and we do not knowingly invite children to provide personal information. If you believe a child has provided information to us, contact tash@neu.ie so we can investigate and take appropriate steps, including deletion where required.

Do not use the app to submit passwords, financial account credentials, government identifiers, medical records or other highly sensitive information. It is not designed as a confidential clinical record system. Sensitive information can also include racial or ethnic origin, religious beliefs, sexuality, genetic information and other categories protected by local law.

If you place that information in a message, memory, document or image, it may be processed as part of your chosen live request. A warning or automated filter cannot guarantee its removal before processing. General Live AI permission should not be treated as permission from another person or as authorization to process every category of protected data. Only provide information about others when you have the necessary permission or other lawful authority.

16. Updates and contact

We may update this policy to reflect changes to our practices, technology or legal obligations. We will revise the effective date and provide additional notice or seek new permission when required for a material change. The app’s privacy controls apply alongside this policy.

Neu Software LLC
Registered in Delaware, United States.

Privacy and support: tash@neu.ie
More help: Support