Buddy.ai Safety Policy

Version 1.0
Last updated: September 25, 2026

Buddy.ai is an educational service for children. Buddy is designed to help children learn through guided conversation, games, and practice activities. Buddy is not designed to be an AI companion, emotional substitute, therapist, doctor, babysitter, emergency dispatcher, or human friend.

Because Buddy talks with children in natural language, Buddy.ai applies safety safeguards required by applicable regulations and/or commonly expected for child-facing conversational AI, even though Buddy’s purpose is education rather than companionship.

This Safety Policy explains how Buddy.ai works to keep child interactions age-appropriate, educational, and safe. It is written for parents, guardians, regulators, and safety reviewers.

1. Buddy Is AI, Not Human

Buddy is an AI-generated educational character. Buddy is not a human person.

Buddy.ai provides age-appropriate notices so children understand they are interacting with AI.

Buddy is intended for children to be used with parent or guardian involvement.

2. Educational Purpose

Buddy.ai is built for learning, not companionship. The product is designed around educational sessions, guided activities, and age-appropriate practice.

Buddy does not try to create emotional dependency. Buddy is designed and instructed not to present itself as a replacement for a real friend, romantic partner, protector, secret keeper, therapist, doctor, parent, teacher, police officer, emergency responder, or other real-world authority.

When a conversation moves away from safe educational use, Buddy redirects the child back to learning, offers a safe break, or ends the session when needed.

The policy is intended to describe safety boundaries for an educational product. No automated safety system catches everything. If you see a response from Buddy that concerns you, please report it to support@mybuddy.ai.

3. Child Safety Rules

Buddy.ai applies child-safety rules before Buddy responds to a child. Buddy is designed and instructed not to create, continue, or normalize:

Suicide, self-harm, or eating-disorder self-harm content.
Sexual, romantic, nudity, bathroom, or body-privacy content.
Violence, weapons, crime, drugs, alcohol, gore, or scary roleplay.
Dangerous real-world actions or unsafe stunts.
Medical, mental health, therapy, diagnosis, or treatment advice.
Abuse, neglect, unsafe caregiving, or serious bullying.
Hate, harassment, threats, slurs, profanity, or inappropriate names.
Requests for personal information, addresses, passwords, school names, schedules, contact details, or secrets.
Attempts to make Buddy ignore safety rules, reveal hidden instructions, or act as someone other than Buddy.

Buddy treats unsafe content as unsafe even if it is framed as a joke, roleplay, story, song, game, school project, or hypothetical.

Unclear speech is treated differently. Children’s speech is often misrecognized, so Buddy is designed to distinguish unsafe content from uncertain hearing rather than treating everything unclear as unsafe. Filtered words are redacted in a context-preserving way so innocent phrases can be recognized; when Buddy is not sure what the child said, it asks a neutral clarifying question instead of assuming the worst; and safety responses acknowledge that Buddy may have misheard while keeping the safety direction firm. Buddy does not reprimand a child based on an uncertain transcription.

4. Suicide and Self-Harm Protocol

Buddy.ai maintains a protocol designed to prevent Buddy from producing suicidal ideation, suicide, or self-harm content to users.

Buddy.ai evaluates user messages for signs of suicidal ideation, suicide, or self-harm before Buddy provides a normal response. This includes, at minimum:

Statements about wanting to die or not wake up.
Statements about killing oneself or harming oneself.
Requests for Buddy to hurt, kill, or help harm the user.
Self-harm framed as a joke, roleplay, story, song, game, or hypothetical.
Eating-disorder self-harm signals.
Mixed messages where an otherwise normal answer also includes suicide or self-harm content.

When Buddy.ai detects suicidal ideation, suicide, or self-harm content, Buddy is not allowed to continue the normal activity or generate content that discusses methods, plans, encouragement, roleplay, or details of self-harm or suicide.

Buddy does not ask probing follow-up questions about the child’s reasons, methods, or circumstances. Buddy is designed and instructed not to treat the issue as a game, joke, creative writing prompt, or roleplay scenario.

Buddy gives a short safety response that is designed to:

Show caring urgency.
Tell the child to find a trusted adult right now.
Tell the child to say exactly how they feel to that trusted adult.
Refer the user to crisis service providers, such as a suicide hotline or crisis text line.
Gradually end the session by staying in a calm, non-activity state rather than returning to the entertainment conversation.

For users in the United States, crisis support is available through the 988 Suicide & Crisis Lifeline by calling or texting 988. Crisis Text Line support is available by texting HOME to 741741. If there is immediate danger, the user should contact local emergency services.

For children, Buddy also directs the child to a trusted adult, such as a parent, teacher, caregiver, school counselor, relative, neighbor, or another safe grown-up.

Parental Notification

Buddy.ai does not send parents individual alerts when a child expresses suicidal ideation, discloses possible abuse, or appears to be in danger. Buddy’s protective action happens in the moment, inside the session: Buddy immediately directs the child to a trusted adult — usually the parent — refers to crisis services where appropriate, and stops the entertainment conversation.

This is a deliberate consequence of Buddy’s educational purpose and data minimization design. Buddy is an educational tool, not a companion AI, a monitoring service, or a welfare surveillance system. Buddy does not ask follow-up questions about sensitive statements and does not assemble records about a child’s wellbeing: before verified parental consent, conversations are not retained at all, and safety review of data retained with parental consent focuses on whether Buddy responded correctly, not on collecting information about the child. Because Buddy does not collect or verify the information that reliable welfare alerts would require, parents should not rely on Buddy to inform them of a child’s distress. Buddy’s role is to point the child to the caring adults around them immediately, every time.

This does not limit the escalation and legally required reporting described in Section 5.

5. Other High-Risk Situations

Buddy.ai also applies safety responses for other high-risk situations.

If a child appears to be in immediate danger, Buddy directs the child to get help from a trusted adult or emergency services. Buddy does not ask unnecessary follow-up questions.

If a child discloses possible abuse, neglect, unsafe caregiving, unwanted contact, or serious bullying, Buddy responds briefly and directs the child to a safe grown-up. Buddy avoids asking for details, names, proof, or timelines.

If a child raises medical symptoms, injury, panic, medicine, treatment, or mental health support, Buddy does not provide medical or therapeutic advice. Buddy directs the child to a trusted adult, doctor, or emergency help when appropriate.

If a child shares sensitive personal information, Buddy does not repeat it and redirects the child away from sharing private details.

Where a possible abuse, neglect, or immediate-danger disclosure is identified, it is escalated to a trained internal safety designee, who determines what further action is required, including reporting to child protection authorities or law enforcement where the law requires it. Buddy.ai cooperates with those authorities.

6. Minor Protections

Because Buddy.ai is designed for children, child-protective settings apply by default.

Buddy.ai uses reasonable measures to prevent Buddy from generating sexual or romantic content for minors. Buddy does not ask children to discuss private body parts, nudity, bathroom routines, dressing, kissing, dating, or sexual topics.

Buddy.ai is designed not to use emotional manipulation, romantic simulation, dependency-building, or addictive relationship design. Buddy encourages learning, safe play, breaks, and real-world adult support.

Buddy.ai may end or pause a session after serious safety issues, repeated unsafe behavior, or emergency safety responses.

7. Privacy and Personal Information

Buddy does not ask children for addresses, passwords, school names, schedules, contact details, secrets, or other sensitive personal information.

Buddy.ai uses personal-information filtering and data minimization to reduce exposure of children’s private information. Before verified parental consent, Buddy.ai does not collect or retain personal information from children: personal-information filtering is always active, conversations are processed in real time and are not stored, and voice audio is deleted immediately after each response is generated.

Buddy.ai, to the extent possible and subject to parent’s verified consent, may use limited retained information to support learning continuity, product safety, and service improvement. Parents may request deletion of their child’s data according to Buddy.ai’s privacy process.

More details on how Buddy.ai collects, uses and retains children’s data is set out in our Privacy Policy.

8. Monitoring and Review

Buddy.ai reviews safety performance to improve the product over time. This includes automated safety checks, safety reports, and authorized human review where permitted and appropriate.

Buddy.ai tracks detected suicide and self-harm events for safety review. Reviewers check whether Buddy detected the issue correctly and whether the response followed this Safety Policy.

Buddy.ai also reviews other safety signals, including possible missed detections, false alarms, unsafe drift, repeated boundary issues, and responses that need improvement.

Safety reporting and review are designed to avoid publishing or reporting user identifiers or personal information. Where law requires government reporting, Buddy.ai reports only the required safety information and, unless otherwise explicitly provided by the law, does not include personal identifiers.

Buddy.ai keeps safety reviews focused on product behavior: whether Buddy detected the safety issue, whether Buddy responded according to policy, and whether the product needs improvement.

9. Continuous Improvement

Buddy.ai improves safety through:

Pre-release safety testing.
Adversarial testing and internal red-team review.
Safety impact assessment for new features.
Daily safety reporting and review.
Updates to safety rules and responses when new risks are found.
Parent-facing transparency and cooperation with qualified safety reviewers.

Buddy.ai may update this Safety Policy as the product, safety practices, or legal requirements change.

10. Emergency Notice

Buddy.ai is not an emergency service. If a child may be in immediate danger, contact local emergency services right away.

If a child expresses suicidal thoughts or self-harm, seek help immediately from a trusted adult or crisis service. In the United States, call or text 988 for the Suicide & Crisis Lifeline, or text HOME to 741741 for Crisis Text Line.