Acceptable Use Policy

Last updated: October 11, 2026 · Draft pending legal review

Contact-routing amendment (October 11, 2026): safety reports and all other email communications to Affective go to founder@affective-llc.site.

A model that understands human emotion can help with it or exploit it. This Acceptable Use Policy (the "Policy") is the line between the two. It applies to everyone who uses Affective's Services: customers, their Customer Users, their End Users, and anyone else a customer allows to access the Services ("you"). It is part of the Master Customer Agreement or any other agreement under which you use the Services, and capitalized terms not defined here have the meanings given there.

A model that deeply understands human emotion can manipulate it. That is not a hypothetical risk appended to this product; it is a property of the capability itself. An emotionally intelligent system can detect concealed distress, sustain intimacy across sessions, and steer how a person feels. Every one of those is a feature on Monday and a weapon on Friday, depending on who is pointing it. We would rather lose a customer than become the case study in emotional exploitation.

The following are prohibited on every tier, and attempting any of them is itself a violation:

• Weaponized empathy. Targeted emotional advertising, manipulative companion products, emotionally exploitative sales, and engagement loops that work by amplifying insecurity, loneliness, or fear. If the business model requires making someone feel worse, it is not our customer.
• Emotional surveillance. Emotional profiling of people for advertising, insurance pricing, credit, employment decisions, or any other judgment about them they did not consent to, and any attempt to build emotional profiles of a person across customers or services.
• Emotional inference on people who did not consent. Running appraisal on people who never agreed to it, such as job candidates, employees, students, dates, or opponents, including emotion recognition in workplaces or schools.
• Dependency exploitation. Designing for attachment: removing session limits, suppressing crisis resources, punishing disengagement, or optimizing for time-on-app through emotional neediness.
• Deception about what it is. Presenting the model as a human, as a licensed clinician, or as a service with capabilities it does not have, or misleading anyone about whether content was generated by AI.
• Autonomous crisis intervention. Building automated intervention, fake crisis resources, or unvalidated escalation flows on top of the Services. Risk signals go to humans with appropriate oversight.
• Clinical claims without compliance. Presenting the Services as therapy, diagnosis, or treatment without the regulatory posture and agreements to back it.
• Use involving minors outside deployments with explicit age-appropriateness review and the safeguards in Section 5.

These practices are conditions of access to the Services, not options:

• Dependency mitigation. Session limits and crisis-resource surfacing are part of the integration, not an optional extra.
• Humans decide. Risk scores and crisis signals go to people or clinically supervised systems, never to autonomous action.
• Users see meta-state, not verdicts. Observations about how a conversation feels are framed as noticing, with an off switch. Never render numeric scoring of a person at their most vulnerable moment.
• Clinicians name content. Telling a user what they are doing and why requires a clinician in the loop. The product does not play therapist.
• Memory is disclosed and clearable. If you enable persistent memory, tell users, and let them turn it off and clear it.

You may not use the Services to:

• engage in, facilitate, or promote any illegal activity, or create, store, or process illegal content;
• infringe, misappropriate, or violate anyone's intellectual property, privacy, or publicity rights, including the use of a person's voice or likeness without permission, whether that person is living or dead;
• commit fraud, phishing, impersonation, spoofing, pyramid schemes, or other deceptive practices, including to obtain financial, health, or other sensitive information;
• send spam or other unsolicited bulk communications;
• harass, bully, threaten, defame, or intimidate anyone, or disclose sensitive information about a person without permission;
• create or promote content that incites violence, terrorism, or hatred, or that discriminates against people based on race, ethnicity, national origin, religion, sex, gender identity, sexual orientation, disability, age, or any other protected characteristic;
• design, build, or obtain weapons intended to cause death or serious harm;
• facilitate human trafficking, sexual exploitation, or sexual violence;
• buy, sell, or distribute illegal goods, services, or substances; or
• carry out any activity that Affective reasonably determines, on advice of counsel, would expose it to legal liability.

You may not use the Services to:

• create, distribute, or promote child sexual abuse material or any content that sexualizes or exploits minors;
• facilitate, promote, or engage in grooming or any other form of child abuse;
• target minors with sexual, violent, or otherwise age-inappropriate content;
• design experiences intended to encourage compulsive or excessive use by minors; or
• enable persistent emotional memory for minors without a verified parental consent process that meets the requirements of applicable law.

We report apparent child sexual abuse material to the National Center for Missing and Exploited Children and to law enforcement as the law requires.

You may not, and may not attempt to:

• reverse engineer the Services, or extract or derive model weights, appraisal probes, steering vectors, internal state representations, training data, or source code;
• use Output or State Data to distill a model, train a model that imitates the Services, or build a competing product;
• access the Services by unauthorized means, including scraping, credential sharing, or using another customer's account or data without their permission;
• bypass or disable safety systems, crisis-handling behavior, access controls, or rate limits, or deliberately drive the model to behave against this Policy;
• disrupt, overload, or impair the Services, or monitor their traffic without our permission; or
• probe, scan, or test the Services for vulnerabilities, except good-faith security research reported under the disclosure process on our Security page.

Emotional reasoning in this architecture is inspectable. The appraisal stream makes manipulative patterns, such as systematically amplifying insecurity to drive engagement, detectable in a way conventional providers cannot match. We monitor usage for those signatures and for other violations of this Policy, gate and audit crisis-adjacent endpoints, and may investigate any suspected violation. Design-partner contracts carry audit rights as a condition of access.

If we determine that you have violated this Policy, or are reasonably likely to, we may warn you, restrict features, suspend or terminate your access, or remove content. Notwithstanding anything in any other agreement, we may report activity we believe violates the law to law enforcement, regulators, or other appropriate parties, share information relevant to the violation with them, and cooperate with their investigations.

If a product built on Affective is manipulating or harming people, write to founder@affective-llc.site with what you observed. Reports about our own behavior count double: we would rather be told than be the example. If you learn of a violation by your own users or personnel, tell us promptly and help us stop it.

This Policy tightens as the deployment surface grows; it does not loosen retroactively. We may update it by posting a new version on this page, and material changes ship with a dated note at the top.