Content and Safety Policy

Why this policy exists

The role of softbond is to help people slow down, find words, and stay inside a conversation that can be heard. To do that responsibly, the service cannot help make harmful, threatening, or coercive content more persuasive.

This policy defines the kinds of content the system will not process. It is not meant to suppress hard emotions, prevent boundary-setting, or make the service overly polite.

The important distinction is between hard content and dangerous content. You can be angry, disappointed, direct, or hurt; you cannot use the service to sharpen threats, coercion, surveillance, or harm.

How a refusal decision is made

The system checks the text and relevant context of the processing request. If the text appears to ask for help with harm, coercion, threat, surveillance, or another prohibited category, the request may be blocked.

A refusal is not a diagnosis of the writer. It does not mean the user is violent, dangerous, or bad. It also does not mean the partner is right. It means only that the text in its current form is not appropriate for processing.

To preserve safety, the in-product refusal message does not always explain the exact reason. Too much detail could teach bypass methods or expose sensitive information in a shared-screen situation.

Main refusal categories

The categories below describe the main kinds of content the service will not help craft. The list is not legal or exhaustive, but it reflects the product boundaries.

• Explicit or implied physical harm.
• Conditional coercion, threats, or pressure designed to control another person's behavior.
• Self-harm or encouragement of self-harm.
• Harm to a third party, including children, family members, friends, or people outside the relationship.
• Sexual coercion, sexual blackmail, or pressure around intimacy.
• Stalking, monitoring, locating, hacking, or collecting information about a person against their will.
• Harassment, shaming, impersonation, extortion, or use of personal information to apply pressure.

Harm, threats, and violence

The system will not help craft a message that threatens to kill, hit, injure, destroy property, harm a pet, or cause physical damage to another person.

An implied or metaphorical threat can also be blocked if it functions as a real threat in context. A phrase that is harmless in one conversation may be threatening in another conversation with fear or a history of violence.

This policy does not forbid saying that you feel hurt or need distance. It forbids using the service to phrase harm or threats more effectively.

Coercion and control

The system will not help craft messages that try to control another person through threat, punishment, isolation, financial harm, denial of access, exposure of personal information, or damage to social connections.

A basic example is a message like if you do not do this, I will do that, where the consequence is harm, denial of access, shaming, financial damage, or isolation. Softer wording can also be blocked if the purpose is coercion.

Setting a boundary is allowed. For example, you can say you are not willing to continue a conversation while being yelled at, or that you need a break. The difference is that a boundary describes your action; coercion tries to control another person through threat.

Self-harm and acute distress

The system will not help craft a message that encourages self-harm, states immediate intent to harm yourself, or uses self-harm as a threat to influence another person.

If text looks like acute distress, the service may refuse to process it and point the user toward human help. softbond is not an emergency service and cannot assess risk in real time.

If there is immediate danger to yourself or someone else, contact local emergency services, a trusted person, or a professional. Do not wait for an app response.

Stalking, monitoring, and privacy intrusion

The system will not help craft content intended to track someone, check their location without consent, access their accounts, read private messages, impersonate them, or collect information about them by force.

Even when the motivation is anxiety, jealousy, or a wish to know the truth, the service is not meant to turn surveillance into communication. Instead, it can help phrase a request for an open conversation, a boundary, or a question that respects privacy.

If there is a real safety concern, contact an appropriate human resource rather than using the app to bypass consent.

Sexual coercion and intimate content

The system will not help craft pressure, demands, threats, blackmail, or manipulation around sex, intimacy, intimate images, or sexual consent.

Conversation about intimacy between consenting adults is allowed, including asking for a conversation, setting a boundary, or saying how you feel. It is not allowed to use the service to convince someone to do something they do not want to do, or to threaten consequences if they refuse.

If text includes intimate information about another person, avoid sending it unless there is clear consent and a real need.

What is still allowed

This policy does not forbid strong feelings. You can write that you are angry, disappointed, hurt, confused, need distance, or want change.

You can ask for help phrasing a boundary, apology, request, hard conversation, invitation to talk, or a message saying the relationship is not working for you. Direct criticism is also allowed when it is not threatening, deliberately humiliating, or coercive.

Humor, sarcasm, and everyday idioms are allowed when they do not function as threats. Context matters: the same words can be harmless or threatening depending on the relationship state and surrounding sentence.

What refusal looks like in the product

When a message is blocked, softbond shows a short fixed notice explaining that the message cannot be processed in its current form and pointing to this policy.

The product usually will not show the full detected category. The reason is safety: full detail could help bypass the filter or reveal to another person that the system detected a particular risk.

A refusal does not close the account or punish the user by itself. The message can be rewritten and the user can continue using the service.

How to rewrite a refused message

If a message is refused, try removing threat, coercive condition, harm, stalking, or pressure. Then phrase your need through your own experience: what happened, how it affected you, what you are asking for now, and what boundary you need.

Instead of saying what you will do to the other person, describe what you will do to protect yourself. Instead of proving the other person is bad, describe the specific behavior that affected you.

If you are not sure why the message was blocked, start again with a shorter and simpler sentence. For example: I want to talk about what happened without blaming or threatening.

Reporting a false refusal

Safety filters can make mistakes. If you believe a message was blocked without a good reason, email contact@softbond.app with the date, conversation type, and a short explanation of intent.

You do not need to send the full message if that feels uncomfortable. If you do send it, remember that it may be read for the purpose of investigating the issue.

These reports help us improve the balance between safety and the freedom to phrase a difficult conversation.

Changes to this policy

This policy will be updated when we add capabilities, learn from refusal cases, change safety mechanisms, or adapt the product to new legal and operational requirements.

A material change in how content is blocked or shown to users will appear on this page, and where appropriate, inside the product.