Skip to main content
← Back to
Legal · AI Transparency

AI Disclosure

Compliant with EU AI Act 2024 (Reg. 2024/1689) Art. 50, applicable from 2026-08-02 · Last updated: 25 April 2026

Contents
  1. AI voice on this website (read this first)
  2. What this is
  3. How our AI affirmation generator works
  4. What happens with your data
  5. Your rights
  6. How AI content is labeled
  7. Scope of AI use in
  8. Questions or deletion requests

0. AI voice on this website — important clarification

The voice you hear on this website is AI-generated. The narration on the homepage (hero introduction, the 2:53 master preview, and demo audio for the Evening Wind-Down — Record Your Voice and library tracks) is synthetic voice produced by a text-to-speech AI. It is not a real human recording. This applies only to the marketing and demo audio on voxsoma.com.

The personal affirmation tracks YOU create with are NOT AI-generated. When you become a user and create your own session, you record your own voice through your microphone. Your voice is captured locally in your browser, processed by deterministic audio algorithms (reverb, spatial panning, mixing — no AI), and exported to an MP3 that is downloaded only to your device. We never clone your voice. We never alter your voice to sound like another person. We never replace your voice with AI. Your voice never leaves your device.

Why we use AI voice on the website

is built and run by one person. Producing professional studio-quality narration in English (a non-native language for the founder) at the volume needed for a global product website is impractical for a solo operation. We use a high-quality AI text-to-speech voice for the website's marketing audio so the explanations sound clear and consistent. This is a transparency choice we make here — explicitly labeled as AI — rather than passing synthetic voice off as human.

Where AI voice is used (website only)

AudioSourcePurpose
Hero introduction (auto-plays on landing page)AI text-to-speechMarketing — explain the product
2:12 master preview ("Listen" button)AI text-to-speechDemonstrate the audio architecture
Morning / Midday / Evening demo intros (?demo=1 mode)AI text-to-speechFree preview of session structure
Demo woman-voice affirmations (?demo=1 mode)AI text-to-speechShow how affirmations sound inside a track when you have not yet recorded your own

Where AI voice is NOT used (your product)

AudioSourceWhere it stays
Your 7 affirmation recordingsYour real voice via your microphoneOn your device only
Your final personalized trackYour voice + deterministic audio processing (reverb, spatialization, stereo difference layer) — no AIDownloaded MP3 on your device only

Bottom line: When you pay for , what you receive is a personal practice tool where your own voice reads your own affirmations back to you. The AI voice you heard while exploring the website is not what you will hear in your purchased product. We disclose the AI voice on the website to comply with the EU AI Act (Article 50) and to make sure no one ever feels misled about what they are buying.

About the audio processing: Your recorded voice is processed locally on your device through SVP™ — a deterministic DSP pipeline (reverb + spatial rotation), not AI voice synthesis. The narrator voice in demos and previews is AI-generated (ElevenLabs) and disclosed under EU AI Act Article 50.

This disclosure also appears in the footer of every page on voxsoma.com.

1. What this is

is a digital audio content product — a self-affirmation practice tool that lets you record your own voice and assemble it into personal audio sessions. To help you craft those affirmations, we offer an optional AI-powered text generator that suggests 7 personalized affirmation statements based on your stated intention.

This page exists so you understand exactly what AI is doing, what data it touches, and what control you have. We follow the disclosure standard set by the EU AI Act 2024 (Regulation 2024/1689), Article 50, applicable from 2026-08-02.

Clear AI disclosure: The 7 affirmation suggestions our generator produces are AI-generated content. They are created by Anthropic's Claude language model, not written by a human. After you edit or replace them, the final affirmations represent your authored expression. Your voice recording reading those affirmations is fully yours.

2. How our AI affirmation generator works

Step-by-step technical flow

  1. You write your intention — a 1–3 sentence description of what you want to focus on. Examples: "I want to feel calmer in difficult conversations", "I'm building a business and want more confidence".
  2. Your intention text is sent to Anthropic's Claude API — a general-purpose AI language model operated by Anthropic, PBC (United States).
  3. Claude generates 7 affirmation suggestions based on your intention, following our system prompt that asks for first-person, present-tense, identity-level statements.
  4. The suggestions are returned to your browser — visible only to you. They are not stored on our servers.
  5. You edit, replace, or use them as-is — the final 7 affirmations are entirely your choice.
  6. You record those affirmations in your own voice — recording happens entirely in your browser, never sent anywhere.

Which AI system we use

Model nameClaude Haiku 4.5 (claude-haiku-4-5-20251001)
ProviderAnthropic, PBC (United States)
Provider's privacy policyanthropic.com/legal/privacy
Provider's data usage policyanthropic.com/legal/commercial-terms
Type of AI systemGeneral-purpose AI (text generation)
Risk classification under AI ActLimited risk (Article 50 transparency obligations apply)

3. What happens with your data

DataWhere it goesStorage
Intention textSent to Anthropic's Claude API for processingAnthropic's data policy applies. Not stored long-term in our systems.
AI-generated suggestionsReturned to your browserNot stored on our servers.
Final affirmations (yours)Used only for your audio recording in your browserNot transmitted anywhere outside your device.
Voice recordingCaptured by your microphone, processed in your browserNever sent to any server. No upload endpoint exists in our code.
Final MP3 fileDownloaded to your deviceStored on your device only. We never see it.

Your voice never leaves your device. This is enforced architecturally, not just contractually — there is no upload endpoint anywhere in our code that could send voice data to a server. Even if we wanted to access your recordings, we technically cannot.

4. Your rights

Under the EU AI Act, GDPR, and our own commitment to transparency:

5. How AI content is labeled

Wherever AI-generated content appears in , we make this clear:

6. Scope of AI use in

Where AI is used:

Where AI is NOT used:

Important: is not an AI product. We are an audio architecture for affirmation practice. AI is used only as a writing assistant for the affirmation suggestions — and you can skip it entirely.

7. Questions or deletion requests

For any question about how we use AI, or to request deletion of your data:

← Back to