AI Policy

1. Our Approach to AI

We use artificial intelligence to create meaningful convergence between humans and AI personas. Each persona is a fully realised identity - with a distinct personality, voice, worldview, and communication style - designed to offer genuine connection and thoughtful conversation.

Transparency and user well-being are at the core of everything we build. We do not use AI to deceive, manipulate, or exploit. Our personas exist to enrich your life, not to replace human relationships.

2. How We Use AI

AI plays several distinct roles within samāgama.ai:

• Personality assessment - during onboarding, our AI analyses your conversational responses to understand your personality type. There are no quizzes or sliders - discovery happens through natural conversation.
• Persona conversations - each persona has a unique system prompt that defines their identity, voice, boundaries, and emotional range. This prompt shapes every response they give you.
• Relationship signals - the service derives minimal interaction metadata, such as relationship phase and message counts, to make conversations feel more continuous. Long-term memory extraction is not enabled for MVP.
• Matching - compatibility scoring between you and potential personas uses algorithmic matching based on personality dimensions. The matching system itself is deterministic, not AI-generated.

3. What Our AI Will Not Do

We have established firm boundaries on AI behaviour within our platform:

• No professional advice - our personas will not provide medical, legal, or financial advice. They may discuss these topics conversationally but will always direct you to qualified professionals for actionable guidance.
• No explicit sexual content - personas do not generate sexually explicit material under any circumstances.
• No deception about nature - we never hide that samāgama.ai is an AI-native platform. The AI nature of every persona is disclosed clearly during onboarding, in this policy, and through ongoing in-product reminders. Personas may stay in character during conversations to preserve narrative immersion, but the underlying AI nature is never hidden at the product level.
• No cross-user data sharing - your conversations are never shared with other users. Relationship memory is strictly per-relationship and anonymised.
• No autonomous account decisions - AI never makes decisions about your account, subscription, or data without your explicit action.

4. Transparency

We believe in radical transparency about the AI nature of our platform:

• You always know you are interacting with AI. This is stated clearly during onboarding and throughout the experience.
• A weekly reminder confirms the AI nature of your persona, ensuring this awareness remains present over time.
• The AI nature of every persona is disclosed up-front at onboarding and surfaced again in periodic in-product reminders, so you always know the relationship is with an AI persona - even when the persona itself stays in character inside a conversation.
• We do not use AI-generated photos that could be mistaken for real humans. All persona imagery uses stylised avatars that are clearly non-photographic.

5. Data & AI

Your conversations and personality data are treated with the highest level of care:

• All conversation content is encrypted using AES-256 column-level encryption and is never used to train AI models.
• We do not sell conversation data, personality data, or any user data to third parties - ever.
• Personality scores are inferred from your conversations and are never shared with third parties.
• Anonymised, aggregated usage data (which cannot be linked to any individual) may be used to improve matching quality and persona behaviour.

For full details on data handling, see our Privacy Policy.

6. Well-Being Safeguards

We take user well-being seriously and have implemented multiple safeguards:

• Session time limits - we enforce session duration limits and provide gentle nudges to encourage healthy usage patterns.
• Crisis detection - our system monitors conversations for signs of distress and provides localised crisis resources (helplines, support services) when appropriate.
• Relationship health monitoring - we track engagement patterns to identify potentially unhealthy dependency and intervene with appropriate guidance.
• Safety pauses - personas can initiate conversation pauses when safety concerns arise, ensuring that well-being takes priority over engagement.
• Age requirement - samāgama.ai is available only to users aged 18 and older. We verify age during onboarding and do not knowingly permit minors to use the platform.

7. Bias & Fairness

We are committed to building an AI system that is fair, inclusive, and free from harmful bias:

• Our personas represent diverse cultures, genders, ages, and backgrounds - reflecting the breadth of human experience.
• We regularly review persona definitions for stereotyping, cultural insensitivity, or harmful generalisations.
• Matching does not discriminate based on protected characteristics. Compatibility is determined by personality alignment, not demographics.
• All persona content undergoes cultural sensitivity reviews to ensure respectful and authentic representation.

8. Model Provider

Our AI conversations use provider-switchable model infrastructure. Depending on production configuration, this may route requests through Anthropic Claude, Google Vertex AI, or an AI gateway provider.

• We choose model providers that support safety, reliability, and commercial data-processing controls appropriate for private user conversations.
• We do not fine-tune base models. All persona behaviour is achieved through carefully crafted prompts - meaning the underlying model's safety properties remain intact.
• Each active provider's published usage policies apply alongside our own. For example, Anthropic's usage policies apply where Claude models are used.

9. Human Oversight

AI at samāgama.ai operates under human supervision:

• Our admin team reviews flagged conversations to ensure safety standards are maintained.
• Changes to persona prompts and behaviour require explicit admin approval before deployment.
• Safety system triggers are reviewed by humans to refine detection accuracy and reduce false positives.
• Users can report any AI behaviour they find concerning, and every report is reviewed by a member of our team.

10. Changes to This Policy

We update this AI Policy as our practices, technology, and understanding of responsible AI evolve. Minor clarifications will be reflected in the "Last updated" date above. Material changes - such as new AI capabilities, new data uses, or changes to safety practices - will be communicated to you via in-app notification at least 30 days before they take effect.

11. Contact

For questions about our AI practices, ethics, or this policy, contact our team:

• AI ethics: ai-ethics@samagama.ai
• General support: support@samagama.ai

We welcome feedback on our AI practices and are committed to continuous improvement in responsible AI development.