Effective Date: October 14, 2026 Last Updated: October 14, 2026
This Privacy Policy explains how SafeTalk ("we," "us," or "our") collects, uses, shares, and retains your personal data when you use our co-parenting communication platform, including our AI-powered Message Coach service.
IMPORTANT: SafeTalk is designed specifically for court-ready communications. Our data retention practices differ significantly from standard consumer applications due to legal and court-readiness requirements.
If you subscribe to the Message Coach service, we additionally collect:
Resolution Maker is a separate paid feature from the Message Coach described in §2.5. It works one arrangement at a time. You tell it what you and your co-parent are stuck on, it writes an arrangement, and you say yes or no. If you say no, you tell it what does not work, and the next arrangement is built from that. If you accept one and give us permission, we send that arrangement to your co-parent, who answers the same way.
If you use Resolution Maker, we collect and create:
What your co-parent sees, and what they do not. Your co-parent never sees the sentence you started with, the reasons you gave for turning an arrangement down, or the AI's working notes. The only thing that reaches them is the text of the arrangement itself, and only after you say yes to sending it. Their reasons are kept from you in the same way.
Where this lives. The session itself — your opening sentence, both of your reasons, the AI's working notes, and the safety reads — is stored on SafeTalk's servers in a place no app can read, yours or your co-parent's. What each of you sees on screen is written to a separate record that only that person can read.
Who at SafeTalk can see it. SafeTalk personnel can access Resolution Maker records for support, safety review, and engineering quality review, and to respond to legal process. As with everything else on SafeTalk, these records can be subpoenaed in a legal proceeding (§4.1, §5.2).
To understand how the app is used and to find and fix problems, SafeTalk collects product-analytics data through two service providers, PostHog and Google Firebase Analytics. This includes which screens you open, which features you use, how long you spend, where you tap, device type, and app version. Through PostHog it also includes "session replay," a video-like playback of how the app's screens looked and how you moved through them. We use it to see what went wrong when something in the app breaks.
Session replay hides every piece of text and every picture on the screen before any replay data leaves your device. The words of your messages, your conversations with the coach, your children's names, and any photo or video shown in the app appear in a replay only as blank shapes. Documents you open from a message open in your phone's own viewer, outside SafeTalk, and are never part of a replay. SafeTalk does not run session replay or PostHog analytics on the screen that shows crisis resources.
PostHog stores this data in the United States. PostHog processes it only as SafeTalk's service provider, under a written agreement and on SafeTalk's instructions, and may not sell it, share it for advertising, or use it for its own purposes. SafeTalk does not sell this data or use it for advertising. Session replays are kept for 30 days and then deleted. Like other SafeTalk records, replay data could be requested in a legal proceeding while it exists.
Product analytics, including masked session replay, is a built-in part of how SafeTalk works and improves, and it is not optional. By creating an account and accepting the Terms of Service, you agree to it. If you do not want it, please do not use SafeTalk. Washington State residents and users in the European Economic Area, the United Kingdom, and Switzerland are treated differently, as described in Sections 6.4(h) and 6.5.
Now and then the app asks a few short questions, such as what is hardest about coparenting for you right now. Answering is up to you. If you answer, we collect what you write, when you wrote it, and which buttons you tap afterward.
We read your answers to learn what to build. We may also use your answer to the first question to show you, inside the app, a feature we think could help, and that screen may quote a few of your own words back to you. The copy of your words made for that screen is deleted when you leave it; we keep a record of which of your words were shown, how they were chosen, and what you tapped.
We never show your answers to your co-parent. They are not part of your co-parent thread, any export, or any court-ready record. Unlike the records described in Section 4, we keep your answers, and the record of what we showed you, for no longer than 12 months, and we delete them sooner if you ask (Section 6.3).
For parents who may live in Washington or Nevada, a person on our team reads the answer first. We use it this way only if it says nothing about anyone's physical or mental health, such as a condition, treatment, or medication. If it does, we don't use it this way.
We process your data based on:
Unlike standard consumer applications, SafeTalk retains communication data indefinitely for legal and court-readiness purposes.
Rationale:
Important: Your coaching conversations (your private messages to the AI coach) are retained alongside your co-parent communications. While coaching conversations are between you and the AI, they may be subject to subpoena or discovery in legal proceedings, including custody disputes. Please keep this in mind as you use the service.
Different SafeTalk products have different deletion postures. Path to Resolution readings (§5.3) are deletable on request because they are personal analytical outputs, not shared communications. The indefinite-retention, deletion-resistant posture in this §4 applies to co-parent chat messages, mediation records, and other shared-communication artifacts; it does not extend to per-user analytical outputs.
When you deactivate your account:
deleted_users collection
with full account historyLegal Basis for Retention After Deactivation:
Under GDPR Article 17 and CCPA, you generally have a right to request data deletion. However, these rights are LIMITED when:
Precedent: This approach is consistent with platforms like OurFamilyWizard and TalkingParents, which have established that court-focused communication platforms may refuse deletion requests due to the nature of the service.
Resolution Maker follows the indefinite-retention posture in this §4, not the Path to Resolution posture. A Resolution Maker session is not a personal analytical output. It is a negotiation between two people that ends in an arrangement one of them sends to the other, and it is tied to a payment record. There is no way to delete a session from inside the product. Requests to delete Resolution Maker data are handled under §4.3 and §6.3.
All communications, mediation data, and shared content are accessible to both connected co-parents. This includes:
Your co-parent retains access even after you deactivate your account.
We may disclose your data to:
We may provide you notice when legally permitted, but are not required to do so.
We share data with categories of third-party service providers:
If you are a Washington resident, you have the right to request a list of the specific third parties with whom we have shared your consumer health data — see Section 6.
The following data is sent to our AI service providers to generate coaching responses and message drafts:
Our AI providers process this data solely to generate responses and do not use it to train their models. Data sent to AI providers is subject to their respective data processing agreements and API terms.
When you request a Path to Resolution reading, SafeTalk reads the chat messages you have exchanged with your co-parent within SafeTalk and sends the following to an AI service provider to generate the reading:
Our AI service provider processes this data solely to generate the reading and is contractually prohibited from using it to train its models, in accordance with our service-provider agreements described below. The reading itself — the four-part text output written about your communication pattern — is stored on SafeTalk's servers under your user account and is visible only to you. Your co-parent generates and views their own reading separately; the two readings are not cross-referenced, and your co-parent does not see your reading.
Path to Resolution readings are treated differently from co-parent chat messages for retention and deletion purposes. Co-parent chat messages are retained indefinitely for court-readiness purposes (see §4.1), and deletion requests for chat messages are generally denied because the messages are shared with another user and may be subject to legal preservation obligations (see §4.3). Path to Resolution readings, by contrast, are personal-to-you analytical outputs — they are not shared communications and they are not part of the court-ready communication record. You may delete your Path to Resolution reading at any time from within the Path to Resolution product surface, and SafeTalk will honor that deletion request on a same-session basis without invoking the §4.3 exceptions that govern chat-message deletion.
Path to Resolution readings sometimes describe patterns in your co-parent's messages as hedged hypotheses (for example, "the late-night messages may be coming from somewhere we can't see"). These are inferences drawn from the chat history you both authored together; they are presented to you for your own reflection and are not shared with your co-parent.
When you use Resolution Maker, the following is sent to an AI service provider:
The provider uses this to write the next arrangement, to check that arrangement against what the two of you have already ruled out, and to run the automated safety read described in §2.6. Our AI service provider processes this data solely to provide the feature and is contractually prohibited from using it to train its models, under the service-provider agreements described below.
The thread messages we send are messages the two of you already wrote to each other, so nothing in them is new to either of you. What you tell Resolution Maker directly is different — your opening sentence and your reasons go to the AI provider, but not to your co-parent (§2.6).
To decide which parents a new feature may fit, and which of their own words to show them, we send an AI service provider:
We do not attach your name, your co-parent's name, or your account ID to it, though the provider will see any names you wrote in the answer yourself. The provider tells us whether the feature may fit and copies out the exact words it would show you, and we check by computer that those words are really in your answer. Our AI service provider processes this data solely for this purpose and is contractually prohibited from using it to train its models, under the service-provider agreements described below.
Resolution Maker is paid. We use Stripe as our payment processor. Stripe is a separate company from our AI service providers and receives none of the data described above.
When you save a card for Resolution Maker, you enter it on a page hosted by Stripe. SafeTalk never sees or stores your card number. Stripe holds the card and gives us two identifiers — one for you as a customer, one for the saved card — which we keep on your SafeTalk account so the card can be charged later without you entering it again.
Nothing is taken when you save the card. A charge is made only if you and your co-parent agree on an arrangement. When we make it, we send Stripe:
We give Stripe your email address for that receipt and for nothing else. We do not store it on Stripe as a contact or marketing field, and we do not use it there for any of our own messages to you.
All third-party providers are contractually obligated to protect your data and use it only for specified purposes.
SafeTalk's AI service providers operate under signed agreements that establish them as "service providers" under the California Consumer Privacy Act and as "processors" under the EU General Data Protection Regulation and equivalent U.S. state and international data-protection laws. Under these agreements, our AI service providers are contractually prohibited from: (a) selling or sharing your data, including for cross-context behavioral advertising; (b) using your data to train, fine-tune, or otherwise improve their AI models; (c) combining your data with personal data they receive from other sources; and (d) processing your data for any purpose other than delivering the service to SafeTalk on SafeTalk's documented instructions.
Our AI service providers are required to: notify SafeTalk of any security incident affecting your data without undue delay; cooperate with SafeTalk's response to your privacy-rights requests; maintain appropriate technical and organizational security measures; flow the same obligations to their sub-processors and remain responsible for sub-processor compliance; and delete or return your data on termination of the relationship. SafeTalk retains audit rights under these agreements. For data subjects in the European Economic Area, the United Kingdom, or Switzerland, cross-border transfers are governed by the Standard Contractual Clauses and applicable UK and Swiss Addenda.
If you are a Washington resident, you have the right to request a list of the specific third parties with whom we have shared your consumer health data — see Section 6 for how to make this request.
These contractual protections do not make SafeTalk a HIPAA-covered entity. SafeTalk is a communications platform with strong service-provider protections; it is not a regulated healthcare service.
We do not sell, rent, or trade your personal information to third parties for marketing purposes.
You have the right to:
Important Limitations: Due to the nature of SafeTalk as a court-ready communication platform, deletion requests will generally be denied when:
You have the right to:
CCPA Exceptions: We may deny deletion requests when necessary for:
Email: info@safetalk-coparent.com Subject: Privacy Rights Request
Include:
Response Time: We will respond within 30 days (GDPR) or 45 days (CCPA).
If you are a resident of Washington State, the Washington My Health My Data Act (RCW 19.373) applies to certain SafeTalk data practices. SafeTalk treats the crisis-detection event log described in §11.2 as "consumer health data" within the meaning of RCW 19.373.010(8), because the trigger categories (self-harm, child safety, intimate-partner violence, threats of violence against an identifiable person) reflect mental-health-adjacent inferences. Other state consumer-health-privacy laws that treat "mental health" data broadly may apply equivalently; SafeTalk operates the practices in this section as the floor for all such laws.
(a) When a record is created. SafeTalk creates a crisis-detection event record only when the automated check shows you crisis resources. The record contains only the items listed in (b).
(b) Purpose, categories, retention. The crisis-detection event log records only: trigger category, the tier of the detection, the resources displayed to you, your continuation choice, and a timestamp. The log does not include a human-reviewer identity, decision, or response-time metadata, because no human review occurs (§11.2). The log is retained on the same indefinite-by-default basis as other SafeTalk records (§4.1), subject to Section 4.3 and applicable statutory exceptions.
(c) No sale; no sharing for cross-context advertising; no sharing for targeted advertising. SafeTalk does not sell crisis-detection event records, does not share them for cross-context behavioral advertising, and does not share them for targeted advertising. The records are processed only to operate the automated detection feature, to maintain SafeTalk's internal calibration (aggregated patterns, not individual user records), and to respond to valid legal process.
(d) AI-provider processing. The pattern-detection step that produces a crisis-detection event runs on SafeTalk's AI providers (see §5.3). Inputs sent to AI providers for this step are the in-session text being evaluated; outputs received are the trigger classification. AI providers are contractually prohibited from training models on this data and are bound by SafeTalk's data-processing agreements.
(e) Your rights. As a Washington resident with respect to crisis-detection event records, you have the rights enumerated in RCW 19.373.040, exercisable through the channel described in §6.3:
(f) Private right of action. RCW 19.373.090 provides a private right of action under the Washington Consumer Protection Act (RCW 19.86) for violations of MHMDA. SafeTalk discloses this exposure here so that Washington residents are on notice of the cause of action available under Washington law. SafeTalk's posture under §11.2 (no human review, no reporting service, event-only logging, the same description of the practice in this Privacy Policy and the Terms of Service) is structured to operate within MHMDA; this disclosure is not a waiver of any defense available to SafeTalk or an admission of any violation.
(g) MHMDA timing. Requests under this §6.4 are responded to within 45 calendar days, the MHMDA statutory window.
(h) Session replay and product analytics — Washington residents. SafeTalk turns session replay off for users it identifies as Washington residents, and does not collect PostHog analytics on the message-thread, coaching, children's-details, or crisis-resource screens for those users. On some iPhones, session replay could restart after the app came back from the background. We are fixing this, and we delete any replays from these users that we find. Washington users still generate the basic data needed to run the app, such as crash and error logs and sign-in events, and basic feature-usage events through Google Firebase Analytics.
SafeTalk turns session replay off for users located in the European Economic Area, the United Kingdom, or Switzerland, and does not treat your acceptance of the Terms as consent to session replay. On some iPhones, session replay could restart after the app came back from the background. We are fixing this, and we delete any replays from these users that we find. Any product analytics for these users is limited to basic usage and stability data, and you may object as described in Section 6.1.
We implement industry-standard security measures:
However, no system is 100% secure. We cannot guarantee absolute security of your data.
Your data may be transferred to and processed in:
We ensure adequate safeguards for international transfers, such as the EU-U.S. Data Privacy Framework (with its UK Extension and the Swiss-U.S. Data Privacy Framework) or Standard Contractual Clauses.
SafeTalk is intended for adults (18+ years). We do not knowingly collect data directly from children under 13 (or 16 in EU). If you believe we have inadvertently collected such data, contact us immediately.
Information about children collected from parents: The Message Coach onboarding process collects your children's first names and ages from you (the parent) to personalize coaching. This information is used solely for the coaching service — to help the AI coach reference your children by name when helping you draft communications. Children's names are never used for advertising, marketing, or any purpose beyond the direct coaching service. Children's names are sent to our AI providers as part of coaching interactions (see Section 5.3).
We may update this Privacy Policy periodically. Changes will be posted on:
Your continued use after changes constitutes acceptance.
SafeTalk uses AI for:
These systems do not make legally binding decisions. All AI recommendations are advisory only. Message drafts require your explicit review and approval before being sent. Final decisions rest with you and your co-parent (or the court).
Message Coach AI Disclosure: When you use the Message Coach, messages sent to your co-parent on your behalf are marked as AI-assisted. Your co-parent will see an indicator that the message was drafted with AI assistance.
This section discloses how SafeTalk's automated crisis-detection feature handles your data.
SafeTalk uses automated detection to surface crisis resources when certain language patterns appear. No SafeTalk staff or contractor reviews these events. Crisis disclosures are not reported by SafeTalk to authorities — users are directed to contact 988, Childhelp, CPS, or 911 directly.
(a) What the system does. SafeTalk's automated pattern detection identifies language in coaching conversations that may indicate (i) imminent risk of self-harm or suicide, (ii) imminent risk of harm to a child, (iii) intimate-partner violence, or (iv) threats of violence against an identifiable person. When such patterns are detected, the system displays direct contact information for category-appropriate crisis resources, may pause or end the coaching thread, and logs the event as an event-only record (§2.5).
(b) What the system does not do. SafeTalk is not a crisis service, mental-health service, mandatory-reporting service, or emergency response service. No SafeTalk employee, contractor, or agent reviews crisis-detection events. SafeTalk does not contact emergency services, child protective services, law enforcement, or any third party on your behalf, and does not notify any other person that a crisis-detection event has occurred. Detection is automated; response is yours.
(c) Resources, not referrals. When a crisis-detection event fires, the system displays direct contact information for services equipped to help. Resource selection is category-specific, not collapsed into a single hotline:
SafeTalk does not contact these services for you. If you or someone else is in danger, you must contact these services directly.
(d) Logging and retention. Crisis-detection events are logged as event records containing only: trigger category, tier of detection, resources displayed, your continuation choice, and timestamp. These records do not include any human-reviewer identity, decision, or response-time metadata, because no human review occurs. The records are retained on the same indefinite-by-default basis as other SafeTalk records (§4.1), are handled on a deletion request as described in Section 4.3, and may be subject to legal process (§5.2). For Washington residents, the records are also subject to §6.4 of this Privacy Policy.
(e) Access scope. Crisis-detection event records sit in SafeTalk's backend-only data layer; they are scoped to your user account at the database-rules layer and are not visible to your co-parent. SafeTalk engineering personnel may access aggregated patterns (not individual user records) for post-launch calibration of the automated detection; individual records are accessed only in response to valid legal process.
(f) No duty assumed. Nothing in this §11.2, in SafeTalk's automated detection, or in the display of crisis resources creates any duty on SafeTalk's part to monitor, warn, report, intervene, rescue, or otherwise act on your behalf or on behalf of any third party. SafeTalk is a communication-coaching tool; automated detection is provided as a safety surface only — it is not a substitute for professional clinical, medical, legal, or emergency services.
Data Protection Inquiries: Email: info@safetalk-coparent.com Website: https://www.safetalk-coparent.com
By using SafeTalk, you acknowledge that you have read and understood this Privacy Policy, including all provisions related to indefinite data retention for legal and court-readiness purposes.
Last Updated: October 14, 2026