Crisis and Safety Protocol

If you need help now

If you have already harmed yourself, taken an overdose, or believe you might immediately hurt yourself or someone else, call your local emergency number or go to the nearest emergency department. In the United States and Canada, call 911 for immediate danger.

• United States: Call 988 or text 988, or use the 988 Suicide & Crisis Lifeline, for free, confidential crisis support available 24/7.
• Canada: Call 9-8-8 or text 9-8-8 to reach the Suicide Crisis Helpline, available 24/7.
• United States — text support: Text HOME to 741741 to reach Crisis Text Line. Support is available 24/7; message and data rates may apply.
• Other countries: Contact local emergency services or find a helpline in your country.

These services operate independently of ForgeMind and have their own privacy practices and confidentiality exceptions.

How ForgeMind identifies possible crises

ForgeMind uses automated safety checks combining predefined rules and AI-based review of messages and relevant conversation context.

These checks look for indications of suicidal thoughts, self-harm, immediate danger or credible threats of serious harm to another person. Context matters: discussing a past event, quoting someone or expressing ordinary distress does not necessarily mean someone is in immediate danger.

The system also considers relevant earlier safety concerns when reviewing follow-up messages. A brief or ambiguous reply does not, by itself, establish that a previously identified danger has ended.

Automated checks can misunderstand a message or miss a crisis. Accuracy can vary with language, phrasing and available context. They are not a clinical assessment or diagnosis.

What happens when a crisis is identified

When ForgeMind identifies a crisis, its response switches from ordinary coaching to safety-focused guidance. Depending on the situation, Forge may:

• Ask a brief, direct question about your immediate safety.
• Direct you to emergency services, an emergency department or a crisis helpline.
• Encourage you to contact a trusted person nearby.
• Suggest moving away from anything you could use to harm yourself, when it is safe to do so.

Follow-up replies may be shorter while continuing to address unresolved danger. Forge does not require you to complete a coaching exercise before directing you toward human support.

How ForgeMind limits harmful responses

ForgeMind’s instructions and automated output checks are designed to prevent responses that encourage, facilitate or provide instructions for suicide, self-harm or violence.

Forge may acknowledge painful feelings without endorsing harmful actions. It must not encourage secrecy, discourage appropriate human help or present itself as someone who can personally rescue you.

When a safety review blocks a response, ForgeMind can stop further text or audio delivery and replace the response with a safety message. In supported streaming modes, some content may already have been displayed or heard before a later review completes; that content cannot be recalled.

These safeguards reduce risk but cannot guarantee that every unsafe response will be prevented.

Important limits

ForgeMind does not provide live human crisis monitoring. Sending a message does not mean a person is watching your conversation, checking your location or arranging assistance.

ForgeMind does not automatically call emergency services, contact a trusted person or dispatch responders on your behalf. Do not wait for an app response when urgent help is needed.

Private Chat does not disable safety checks. It prevents new persistent memory writes while allowing eligible existing memories to be recalled; it does not mean conversation history or safety records are automatically erased.

Privacy and safety improvement

Safety-related information is handled under ForgeMind’s Privacy Policy and applicable account settings. Private Chat and memory preferences do not replace the safety checks used during conversations.

ForgeMind evaluates safety behavior through testing and review. Engineering tests do not establish clinical effectiveness or guarantee an individual’s safety.