Consent Document

Why consent is required

softbond handles personal and sensitive text. Before the service suggests wording, a question, a summary, or conversation structure, users should understand what happens behind the scenes rather than discover it after a conversation has started.

Consent clarifies the product boundaries: what is processed, what is not shared with a partner, how approval before sharing works, and what the risks and limits are for content generated by an intelligent system.

The goal is not to flood users with dialogs. The goal is one transparent moment before meaningful use, followed by easy access to this document at any time.

When consent is shown

Consent is shown before first use of softbond AI-powered features or before first entry into a conversation where personal text may be processed.

If we later add a feature that performs a new kind of processing or shares information differently, we may show additional consent or a change notice before continued use.

Continuing after approval lets the product operate the capabilities described here, subject to the Privacy Policy and Terms of Use.

You are asking softbond AI for help

softbond AI can help phrase, shorten, soften, clarify a boundary, ask a question, summarize a conversation, suggest a possible response, or structure a Repair Session.

softbond AI is not a person, therapist, mediator, lawyer, or someone who knows what the other person truly thinks. It works from the text and context provided to it.

The user remains the person who decides whether to accept the suggestion, edit it, delete it, or send it. Nothing is sent automatically because a response was generated.

Relevant text is processed

When you use a softbond AI feature, the text needed to operate that feature may be processed through the OpenAI API. This may include a message you wrote, a change request, parts of conversation context, or a received response.

The system should send only the context needed. Not every piece of content you wrote in the product is sent for every action, and not every private conversation is exposed for every wording request.

During OpenAI processing, the relevant text is read by the processing provider in order to generate a response. This is part of the technical operation of the feature, which is why consent is required before use.

Private and shared stay separate

A private conversation is a place to think before speaking. Your partner does not see what is written there, drafts that were created, rejected wording, or change requests that were not chosen for sharing.

A Repair Session is a shared space, but even there the transfer is not automatic. A message crosses only after explicit approval by the person who wrote it.

Inviting a partner does not turn personal history into shared history. Only content approved for sharing or created inside the shared space is treated as shared.

Approval before a message or agreement

When softbond suggests wording, the user can edit it, ask for another version, copy it, send it, or discard it. Choosing one of those options is separate from the processing itself.

A Living Agreement is created only when users choose to turn an understanding or decision into a shared agreement. It should not be created behind the scenes from unapproved text.

The simple principle is that content does not move from a personal place to a shared place without a clear user action.

Derived insights and My Data

softbond may store derived insights, such as recurring topics, wording preferences, active agreements, or reminders the user requested. These insights are meant to help understand the conversation, not diagnose a person.

The My Data area is intended to show what is stored, what has been derived, and what can be exported or deleted. Users should not have to guess what data exists.

Not every system record is shown as an everyday insight. Security records, logs, and operational records may be kept for security, support, or legal reasons.

Safety filter

Every message that goes through a softbond AI feature may be checked to avoid helping craft content that may harm, threaten, coerce, stalk, or encourage harm.

The filter applies to specific text, not to the user's identity. A refusal is not a diagnosis, not a conclusion that the user is violent, and not a conclusion about the relationship.

When a message is refused, it can be rewritten without threat, coercion, or harm. If there is distress or danger, seek appropriate human help.

No therapy and no emergency service

softbond is not couples therapy, mental-health counseling, a hotline, or emergency support. It can help with wording and organizing thought, but it does not replace a professional.

If there is violence, threat, control, fear, danger to children, suicidality, or severe distress, do not rely on the app. Contact emergency services, a professional, or a trusted person who can help.

The system does not know the full relationship history and cannot assess risk like a trained human professional familiar with the situation.

Security and encryption

Content and service data are stored and transmitted encrypted according to strict standards. Separation between personal information and shared couple information is part of the core design.

At the same time, using softbond AI features requires processing of the relevant text by the processing provider. It is important to distinguish encryption in storage and transit from a processing action where the system must read text to generate an answer.

Human access to content should be limited to situations with a clear need, such as technical support, security, or legal obligation, and recorded under service procedures.

Independent account for each user

Even inside a relationship, each user keeps an independent account. Inviting a partner does not give them control over your account or access to personal content.

Each user can request export, deletion, or account closure for their own information. Shared information is handled carefully so each person can leave without necessarily deleting history the other person is entitled to see.

The principle is autonomy for both sides. softbond should not become a tool for one person to manage the other.

Withdrawing consent

You can stop using softbond AI features, request data deletion, or close your account. These requests are handled under the Privacy Policy and Terms of Use.

Withdrawing consent may limit features that depend on text processing, such as rephrasing, summarization, or rehearsing a possible response. Without that processing, parts of the experience cannot work the same way.

Withdrawing consent does not necessarily undo processing that already occurred before withdrawal, but it will affect future use where the product allows.

Changes to consent

If we add a new kind of processing, change processing providers, expand use of derived insights, or change the separation between private and shared spaces, we will update this document and show an appropriate notice.

A material change should not happen silently. Active users will receive a clear explanation before continuing to use a capability that changed significantly.