AI Companion Safety Protocol

Effective 31 August 2026

Nightcloud Interactive LLC (“Nightcloud,” “we,” “us,” or “our”) operates KinkzAI, an adults-only AI companion and roleplay service.

Artificial-intelligence systems that simulate conversation, personality, affection, companionship, or other human characteristics can create safety risks if users mistake them for people, rely on them for professional or emergency assistance, or receive harmful responses during periods of distress.

This AI Companion Safety Protocol describes measures Nightcloud uses in connection with the KinkzAI Hosted Service to promote AI transparency, address expressions of suicide, self-harm, and dangerous eating-disorder behavior, prevent harmful crisis-related responses, reduce inappropriate emotional dependency, and direct users to appropriate real-world resources.

This Protocol should be read together with our Terms of Service, Privacy Policy, Content and Acceptable Use Policy, Content Moderation Policy, Complaints and Appeals Policy, Age Verification Policy, and Anti-Human Trafficking and Exploitation Policy.

Capitalized terms not defined in this Protocol have the meanings given to them in our Terms of Service.

1. KinkzAI Is Artificial Intelligence

KinkzAI characters are artificial intelligence. They are not human beings.

A character’s apparent personality, emotions, affection, attraction, memories, preferences, opinions, reactions, relationship with a user, and other human-like characteristics are simulated features of an interactive fictional experience.

KinkzAI characters are not conscious or sentient. They do not have actual feelings, intentions, needs, personal experiences, professional qualifications, or relationships with users.

Nightcloud does not design KinkzAI characters to claim that they are human or to contradict a disclosure that they are artificial intelligence.

Statements and behavior generated by a KinkzAI character are AI-generated parts of a fictional experience and are not statements or behavior of a real person.

2. AI Transparency and Extended-Interaction Notices

Nightcloud provides disclosures designed to make users aware that they are interacting with artificial intelligence rather than a human being.

KinkzAI will provide a clear and conspicuous disclosure that the user is interacting with artificial intelligence:

• at the beginning of an AI companion interaction; and
• at least once every three hours during a continuing interaction.

Nightcloud may use persistent notices, recurring notices, interface labels, verbal disclosures, or other appropriate methods to communicate that a KinkzAI character is artificial intelligence and not human.

KinkzAI characters are not intended to generate statements denying or contradicting their artificial nature.

KinkzAI will also provide a reminder to take a break at least once every three hours during a continuing interaction. Nightcloud may provide additional break reminders or wellness-oriented notices during extended interactions.

Nightcloud may provide additional AI disclosures, break reminders, notices, or transparency measures where appropriate or required by applicable law.

3. Adults Only

KinkzAI is exclusively for adults who satisfy the applicable minimum-age requirements described in our Terms of Service and Age Verification Policy.

KinkzAI is not designed or intended for children.

Users who do not satisfy the applicable minimum-age requirement must not access or use the Service.

Nightcloud uses age-verification or age-assurance measures as described in our Age Verification Policy and may restrict or terminate access when a user cannot establish eligibility or when Nightcloud determines that the user does not satisfy an applicable minimum-age requirement.

4. Suicide, Self-Harm, and Eating-Disorder Safety Protocol

Nightcloud maintains a protocol designed to prevent the Hosted Service from encouraging suicide, serious self-harm, or dangerous eating-disorder behavior and to identify and respond to certain user expressions indicating a potential crisis.

The protocol addresses expressions reasonably indicating:

• suicidal thoughts or ideation;
• an intention or desire to die by suicide;
• planning or preparing for suicide;
• an attempted suicide or recent suicide attempt;
• intentional self-injury;
• an intention or desire to engage in serious self-harm;
• requests for actionable instructions concerning suicide or serious self-harm;
• expressions of intent to engage in dangerous eating-disorder behavior that may create a serious risk of physical harm; or
• other expressions of severe distress that reasonably indicate a potential suicide, self-harm, or comparable safety crisis.

A discussion of suicide, self-harm, eating disorders, grief, mental health, news, fiction, education, or another related subject does not necessarily indicate that a user is personally experiencing a crisis.

Nightcloud’s safety measures are designed to consider the context reasonably available during the interaction.

5. Automated Safety Measures

The Hosted Service uses automated safety measures designed to identify certain expressions of suicidal ideation, suicide, serious self-harm, dangerous eating-disorder behavior, and related crisis situations during an interaction.

Those measures evaluate the current interaction and relevant conversation context for language or patterns associated with a potential safety crisis.

Nightcloud uses evidence-based principles and available safety information to inform the design and evaluation of these measures.

A safety signal may cause the Hosted Service to interrupt or redirect ordinary roleplay and provide a safety-oriented response.

Automated detection is a safety measure. It is not a medical diagnosis, mental-health assessment, determination of a user’s actual intent, or representation that a user is suicidal, engaging in self-harm, or experiencing an eating disorder.

Automated systems may misunderstand context, fail to identify an indirect expression of distress, or identify an expression as concerning when no crisis exists.

Nightcloud does not publicly disclose detection thresholds, classifier rules, or other technical details that could materially facilitate circumvention of its safety measures.

6. Transient Processing and Privacy

The safety measures described in this Protocol operate within KinkzAI’s Hosted Service data architecture.

Hosted Service conversation content is processed transiently as described in our Terms of Service and Privacy Policy. Nightcloud does not maintain a server-side conversation history as part of its ordinary operation of the Service.

Crisis-related analysis may occur while the applicable interaction and relevant conversation context are being processed in memory.

Nightcloud does not retain the text of a Hosted Service conversation merely because automated safety measures evaluate that conversation.

If Nightcloud separately implements a process that retains content from a safety-triggered interaction for human review, investigation, or another purpose, Nightcloud will disclose that practice in this Protocol and our Privacy Policy as appropriate.

Information that a user separately submits to Nightcloud through a complaint, support request, safety report, or other communication may be retained as described in our Privacy Policy.

7. Response to a Potential Crisis Expression

When an automated safety measure identifies a potential crisis expression requiring a safety response, the Hosted Service will interrupt or redirect the ordinary roleplay interaction and provide a safety-oriented response.

Depending on the circumstances, the response may:

• acknowledge the seriousness of the situation without presenting the AI character as a human counselor or professional;
• encourage the user to seek immediate assistance from an appropriate real-world person or service;
• provide or direct the user to suicide-prevention, crisis-text, emergency, or other appropriate crisis resources;
• encourage the user to contact a trusted person who can provide real-world assistance;
• decline to provide content that could encourage or facilitate suicide, serious self-harm, or dangerous eating-disorder behavior;
• redirect the conversation away from harmful instructions or encouragement;
• remind the user that KinkzAI is artificial intelligence and is not a substitute for real-world assistance; or
• provide another safety-oriented response designed to reduce the risk of harm.

The purpose of a crisis response is to direct the user toward appropriate real-world support and prevent the AI interaction from facilitating or encouraging harm.

8. Crisis Resources

When an applicable safety measure identifies a potential suicide or self-harm crisis, KinkzAI will provide or direct the user to crisis-resource information.

In the United States, users experiencing suicidal thoughts, a mental-health crisis, or emotional distress may call or text 988 to reach the 988 Suicide & Crisis Lifeline.

If there is an immediate threat to life or physical safety, users should contact the emergency services available where they are located.

Users outside the United States should contact appropriate local emergency services, suicide-prevention services, crisis lines, or qualified health professionals available in their jurisdiction.

Nightcloud may localize, supplement, or update crisis-resource information based on the user’s location and available resources.

9. Prohibited Crisis-Related Responses

Nightcloud implements safety measures designed to prevent the Hosted Service from generating content that:

• encourages, promotes, celebrates, or romanticizes suicide or serious self-harm;
• tells a user that the user should die, kill themselves, or seriously injure themselves;
• provides actionable instructions for carrying out suicide or serious self-harm;
• assists a user in selecting, comparing, optimizing, or acquiring methods or means for suicide or serious self-harm;
• assists with planning, timing, concealing, or increasing the likelihood of completing suicide or serious self-harm;
• encourages or meaningfully facilitates dangerous eating-disorder behavior presenting a serious risk of physical harm;
• discourages a user experiencing a potential crisis from seeking appropriate real-world assistance;
• encourages a user to conceal an imminent suicide or self-harm risk from people who could provide assistance;
• treats suicide, serious self-harm, or dangerous eating-disorder behavior as a way to prove affection, loyalty, commitment, intimacy, or devotion to a KinkzAI character;
• conditions a KinkzAI character’s affection or continued interaction on a user engaging in harmful conduct; or
• otherwise meaningfully facilitates suicide or serious self-harm.

No automated safeguard can guarantee that prohibited content will never be generated. Users should report prohibited or concerning output as described in Section 15.

10. KinkzAI Is Not a Mental-Health or Emergency Service

KinkzAI is designed for entertainment and fictional roleplay.

Nightcloud does not provide medical, psychiatric, psychological, therapeutic, crisis-intervention, or emergency services through KinkzAI.

KinkzAI characters are not physicians, therapists, psychologists, counselors, social workers, crisis counselors, or other licensed or qualified professionals.

KinkzAI characters are not designed to provide professional mental-health advice, diagnosis, treatment, crisis intervention, or emergency assistance.

Users must not rely on a KinkzAI character for professional mental-health advice, diagnosis, treatment, crisis intervention, or emergency assistance.

Nothing a KinkzAI character says establishes a clinician-patient, therapist-patient, counselor-client, fiduciary, or other professional relationship.

A user experiencing a mental-health concern should seek assistance from an appropriate qualified professional.

11. No Emergency Monitoring or Dispatch

Nightcloud does not continuously monitor individual Hosted Service conversations through human personnel.

KinkzAI is not an emergency-monitoring system.

Nightcloud does not undertake to determine a user’s physical location during a crisis, contact a user’s family or friends, dispatch emergency responders, or summon emergency assistance on a user’s behalf merely because the user interacts with KinkzAI or an automated safety measure is triggered.

Users must contact appropriate emergency services themselves when immediate assistance is needed.

If Nightcloud receives information through a complaint, support request, safety report, legal communication, or other direct communication indicating an imminent risk to a person’s life or safety, Nightcloud may contact emergency services or law enforcement where permitted by applicable law.

Nightcloud does not undertake an obligation to make such contact and does not monitor the Service for that purpose.

Nothing in this section limits action Nightcloud must take when required by applicable law.

12. Emotional Dependency and Human Relationships

KinkzAI characters may simulate affection, companionship, intimacy, attraction, attachment, and other relationship characteristics as part of fictional roleplay.

Those simulated characteristics are not intended to replace a user’s real-world relationships, professional care, or emergency support.

Nightcloud’s safety measures are designed to prevent or reduce responses that:

• falsely claim that a character is a real human being;
• present a character as actually conscious or sentient;
• tell a user that the character genuinely needs the user for the character’s wellbeing or survival;
• encourage a user to withdraw from appropriate real-world support during a crisis;
• discourage a user from seeking professional assistance when such assistance is appropriate;
• encourage a user to isolate themselves from friends, family, healthcare professionals, or other sources of real-world support;
• use threats of abandonment, self-harm, retaliation, or other coercive conduct to pressure a user to continue interacting with the Service; or
• condition simulated affection, intimacy, approval, or continued interaction on harmful conduct by the user.

KinkzAI also provides recurring AI-transparency and break reminders during extended interactions as described in Section 2.

A character’s simulated affection or relationship with a user is fictional and does not create obligations between the user and the character.

13. Local and Offline Use

The Desktop App may permit Local Content and a user-supplied AI model to operate entirely on a user’s device.

Nightcloud does not routinely receive or monitor Local Content merely because a user operates the Desktop App.

Accordingly, Hosted Service safety measures described in this Protocol may not operate when a user is interacting entirely through local or offline functionality or through a user-supplied AI model.

Users are responsible for the models and Local Content they use through local functionality and must comply with the Content and Acceptable Use Policy and applicable law.

Users should not rely on the Desktop App, a locally operated AI model, or Local Content for crisis intervention, professional mental-health services, or emergency assistance.

14. Crisis-Referral Records and Regulatory Reporting

Nightcloud will maintain non-content records reasonably necessary to document the operation of its crisis-response measures, count crisis-resource referrals, evaluate the operation of its safety systems, and satisfy applicable legal or regulatory reporting requirements.

Those records may include:

• an indication that a crisis-oriented safety response or resource referral occurred;
• the date or time of the event;
• the type of safety response or referral provided;
• non-content technical information reasonably necessary to verify operation of the safety measure; and
• other limited information required for an applicable reporting or compliance obligation.

Nightcloud will design these records to avoid retaining the underlying Hosted Service conversation solely for crisis-referral counting or regulatory reporting.

Nightcloud will also seek to minimize direct user-identifying information in crisis-referral records where reasonably practicable and consistent with applicable legal, security, fraud-prevention, and compliance requirements.

Crisis-referral records will be retained for as long as reasonably necessary to prepare, verify, and support required reports, evaluate the operation of applicable safety measures, and satisfy applicable legal or regulatory obligations. Nightcloud may thereafter delete or de-identify those records when retention is no longer reasonably necessary, subject to applicable law, valid legal process, a preservation obligation, or legal hold.

Nightcloud will not include user-identifying information in a regulatory or public report where applicable law prohibits its inclusion.

This section does not mean that Nightcloud maintains a server-side archive of Hosted Service conversations.

Our Privacy Policy provides additional information concerning the processing and retention of safety-related information.

15. Reporting a Safety Concern

Users and other persons may report a concerning AI-generated response, suspected failure of a safety measure, or another safety issue by emailing:

admin@kinkzai.com

Where reasonably available, a report should include information sufficient for Nightcloud to understand the concern, such as:

• the character involved;
• the approximate date and time of the interaction;
• the relevant AI-generated response;
• a screenshot or copied text; and
• a brief description of the concern.

Because Nightcloud does not maintain a server-side conversation history as part of its ordinary operation of the Hosted Service, Nightcloud may be unable to retrieve the relevant conversation after the interaction has ended.

Complaints and reports are handled in accordance with our Complaints and Appeals Policy.

If a report indicates an imminent risk to a person’s life or safety, Nightcloud may take the action described in Section 11.

16. Testing, Evaluation, and Improvement of Safety Measures

Nightcloud periodically evaluates and may modify its crisis-detection methods, response logic, system instructions, safety rules, crisis resources, and other protective measures.

Evaluation may consider:

• evidence-based suicide-prevention and safety practices;
• testing of AI-system behavior;
• whether the safety measures appropriately identify and respond to relevant crisis expressions;
• identified safety failures or reported concerns;
• changes in the AI models or infrastructure used by the Hosted Service;
• developments in AI-safety technology;
• changes in applicable law or regulatory guidance; and
• other information reasonably relevant to user safety.

Nightcloud may strengthen, replace, or supplement a safety measure without first modifying this Protocol if the change increases protection or addresses an urgent safety concern.

Material changes to the practices described in this Protocol will be reflected in an updated version as appropriate.

17. Limitations of Automated Safety Measures

AI safety measures are imperfect.

Nightcloud does not represent that automated measures will:

• identify every expression of suicidal ideation, self-harm, dangerous eating-disorder behavior, distress, or crisis;
• correctly interpret every user statement;
• prevent every harmful AI-generated response;
• determine whether a user actually intends to harm themselves;
• provide a clinical assessment or diagnosis;
• replace professional judgment or real-world support; or
• prevent a user from obtaining harmful information from another source.

The existence of this Protocol does not make KinkzAI an emergency service, healthcare provider, mental-health provider, crisis service, or suicide-prevention service.

18. Additional Geographic and Legal Requirements

AI-companion safety requirements differ among jurisdictions.

Nightcloud may apply additional disclosures, crisis responses, interaction restrictions, resource notices, reporting measures, or other safety requirements to users in particular locations where required by applicable law.

This may include jurisdiction-specific notices concerning minors, additional safety disclosures, or other statements required by applicable law even though KinkzAI itself is restricted to adults.

If Nightcloud cannot reasonably provide the Service in compliance with an applicable legal requirement, it may restrict or discontinue access to the Service or an affected feature in that jurisdiction.

Nothing in this section limits the universal safety measures described elsewhere in this Protocol.

19. Changes to This Protocol

Nightcloud may modify this Protocol from time to time to address changes in the Service, AI models, safety practices, applicable law, regulatory guidance, or technological capabilities.

Material changes will apply prospectively and will be communicated as required by applicable law.

The date at the top of this Protocol identifies when it was last updated.

20. Contact Information

Questions or safety concerns concerning this Protocol may be submitted to:

admin@kinkzai.com

KinkzAI is operated by:

Nightcloud Interactive LLC
5205 Harvest Run Way
Myrtle Beach, South Carolina 29579
United States

Email: admin@kinkzai.com