AI Transparency

Introduction

This document sets out the public commitment of Somia™, an emotional wellbeing project operated by Yeshcube Tech, S.L., to ensure a clear, honest and understandable relationship between people and the artificial intelligence used in our applications, devices and integrated experiences.

For Somia Bloom services, this commitment applies to experiences available at https://bloom.somia.io.

Transparency is one of Somia's core ethical pillars and part of our commitment to human-centric design, digital non-maleficence, scientific integrity and respect for user autonomy.

This Commitment is binding for the design, development and communication of all technology created within the Somia ecosystem.

We believe that artificial intelligence should be understandable, not mysterious. Useful, not invasive. And always in service of human wellbeing, never above it.

1. Transparency about the nature of the system

1.1 What Somia is, and what it is not

From the first use, Somia clearly informs that:

You are interacting with artificial intelligence, not with a human being.

The AI operates through conversational models and generative analysis systems designed for emotional wellbeing.

There are no hidden human operators or human intervention in your conversations.

The AI can make mistakes, omit information or only partially understand your context.

1.2 How Somia's conversational AI works

Somia uses advanced conversational AI models that process your voice or text messages to:

Understand the emotional content and context of what you share

Generate empathetic, validating and wellbeing-oriented responses

Personalize the experience according to your preferences and memories (if you enable them)

Detect language patterns that may indicate emotional distress

These models do not think or feel. They process linguistic patterns through advanced statistical analysis. They have no consciousness, intention or moral judgement.

1.3 Why it is important that you know this

Understanding that you are speaking with an AI allows you to:

Set realistic expectations

Protect your autonomy and decision-making

Know when to seek help from a human professional

Understand the technical and ethical limits of the system

Use Somia in a more conscious and healthy way

Somia will never pretend to be something it is not. This transparency is the foundation of a genuine relationship of trust.

2. Purpose and limits of Somia's AI

2.1 What Somia is designed for

Somia is an artificial intelligence specifically designed for:

Preventive emotional wellbeing: helping you manage everyday emotions, stress and self-care

Empathetic conversational support: offering you a safe, nonjudgmental space to express yourself

Wellbeing personalization: adapting the experience to your rhythms, preferences and emotional needs

Healthier habits: helping you identify patterns, establish routines and strengthen self-care practices

2.2 What Somia is NOT designed for

Somia is not, does not replace, and does not intend to replace:

Professional psychological care

Clinical diagnosis of mental disorders

Psychiatric or psychotherapeutic treatment

Emergency services or crisis intervention

Medical or pharmacological advice

2.3 Technical limits of the AI

Somia's AI:

Uses linguistic patterns to generate responses, not real emotional understanding

Analyses a limited context from previous conversations (depending on memory settings)

Has no access to external information about you beyond what you directly share

Cannot verify objective facts about your life, health or environment

Does not make clinical judgements or assess symptom severity

Does not automatically contact emergency services

2.4 When to seek professional help

You should seek human professional assistance if:

You experience thoughts of self-harm or suicide

Your distress severely affects daily functioning

You need a clinical diagnosis or treatment

You are going through acute emotional crisis

You require structured therapeutic support

Somia does not understand or fails to respond appropriately

Somia can accompany you, but it cannot replace the irreplaceable value of a human professional.

3. Transparency about memories and personalization

3.1 What memories are in Somia

Memories are small pieces of information that Somia may remember between conversations to personalize your wellbeing experience.

Unlike other AI systems, in Somia:

You decide whether memories exist or not

You control which categories of information are remembered

You can view, edit or delete any memory at any time

Memories are stored primarily on your device, not on remote servers

3.3 Memory categories: the 18 blocks

Somia organizes memories into 18 blocks across 4 wellbeing dimensions:

I. Self-knowledge (5 blocks)

Personal preferences: likes, interests, hobbies

Energy rhythms: moments of lower or higher vitality

Communication style: how you prefer to express yourself and receive support

Daily routines: structure of your day

Sleep and rest: sleep patterns and quality of rest

II. Emotional wellbeing (5 blocks)

Stress factors: situations, contexts or people that generate tension

Sources of joy: activities or people that bring you wellbeing

Emotional patterns: recurring tendencies in how you experience emotions

Self-care signals: indicators that you are taking care of yourself

Wellbeing activities: practices that help you feel better

III. Personal development (4 blocks)

Personal goals: aspirations and objectives

Professional challenges: work-related or vocational difficulties

Learning themes: areas you are exploring or wish to explore

Achievements and progress: recognition of milestones

IV. Environment and connections (4 blocks)

Social connections: meaningful relationships and support networks

Environmental impact: how different environments affect you

Sources of support: people or resources you rely on

Inspirations: values, ideas or references that motivate you

3.4 Memories in the Somia Bloom pilot

In the pilot at the Universitat de València Science Park (https://pcuv.somia.io), Somia keeps, only with your permission and on our server, three kinds of short notes: what describes your work and how you organize yourself, topics left open and a short summary of your latest conversations. The 18 category blocks are not used.

Each experience receives only part of them: the guided experiences, at most two notes related to their purpose; the open conversation, a wider but limited context. What you state carries more weight than Somia's notes. You can review, correct or delete everything in “My area”, or ask Somia during the conversation.

4. Transparency about data use in AI

Somia processes your data with the highest security and privacy guarantees. For detailed information on:

Which data we process

Which data stays on your device and which is processed on servers

Technology partners that process your data

Retention periods

Your rights regarding your data

Please consult our Privacy Policy.

5. Transparency about emotional safety

5.1 Emotional risk detection systems

Somia incorporates detection mechanisms designed to identify language that may indicate:

Self-harm or suicidal thoughts

Acute emotional crisis

Severe distress or hopelessness

Personal safety risks

5.2 Critical limitations of these systems

It is essential that you understand:

These systems are not perfect: they may fail to detect real risk

Not all crises are detectable through language

They do not provide clinical intervention: they only redirect to professional resources

They do not contact emergency services: no automatic activation of external help

They do not replace professional support: they are preventive support, not clinical safety systems

5.3 Professional help resources

If you are in crisis or need urgent help:

Spain:

Teléfono de la Esperanza: 717 003 717 (24h)

National Suicide Hotline: 024 (24h, free & confidential)

Emergencies: 112

International:

Find global resources at https://findahelpline.com

14. Executive Summary: What you must remember

If you only read one section, make it this one:

Somia is AI, not a human

You are speaking with artificial intelligence designed for emotional wellbeing, not a psychologist or therapist.

Somia has clear limits

It does not diagnose, treat or replace professional help. In crisis, seek human assistance.

You control your memories

You decide what is remembered, stored or synced. Everything is optional and reversible.

Your data stays on your device

Local-first architecture: your memories live on your phone/computer. They only go to the cloud if you enable it.

We use third-party AI

ElevenLabs for voice and leading partner models for conversation. We require strict ethical and privacy compliance.

AI can make mistakes

Errors and misunderstandings happen. Do not rely blindly. Keep your judgement.

Permanent transparency

This commitment is binding. We uphold it and update it. If we fail, you have the right to demand accountability.

15. Final Commitment

This is not a defensive legal document. It is a public promise.

We commit to making Somia:

Understandable: you will always know what it does and what it does not do

Respectful: your autonomy and privacy are non-negotiable

Honest: we will not hide limitations or risks

Safe: your emotional wellbeing comes before any business metric

Yours: your data belongs to you, not to us

If we ever fail to uphold this commitment, you have the right to hold us accountable.

Thank you for trusting Somia.

Contact

Yeshcube Tech, S.L.

C. de la Travesía, SN

46024 Valencia, Spain

General email: hello@yeshcube.com

Legal email: legal@yeshcube.com

Document in force since: November 15, 2025

Last update: November 15, 2025

Next scheduled review: November 2026

Version: 1.0