AI-Powered Legal Evidence Documentation for Victims of Abuse

SafeVoice is a voice-first AI companion designed to help victims of abuse safely document incidents, preserve evidence, and prepare for legal action — without typing, legal expertise, or putting themselves at risk.

The app is built with safety-first principles, including a decoy interface, PIN-protected access, and instant exit mechanisms, making it usable even in high-risk environments.

Most victims of abuse never report incidents.

Not because they do not want help — but because:

• Documenting evidence is unsafe
• Trauma makes writing and recall difficult
• Victims do not know what details are legally important
• Written journals can be discovered by an abuser
• Legal support is expensive and inaccessible

As a result, evidence is lost, escalation goes unnoticed, and victims remain trapped.

SafeVoice provides a private, secure space where victims can simply speak — and the system handles the rest.

It works quietly, looks harmless, and only reveals its real functionality when intentionally unlocked by the user.

Introduces SafeVoice and its purpose.

This is the default interface shown when the app is opened.
It appears as a harmless meditation or wellness app.

A real guided meditation experience powered by an AI meditation agent using ElevenLabs.
This helps users regulate emotions while maintaining the decoy illusion.

The real SafeVoice flow unlocks only after a deliberate long-press followed by PIN verification.

Main dashboard after unlocking, providing access to documentation and history.

Quick exit and safety-first controls designed for high-risk situations.

Chronological, secure evidence timeline showing documented incidents over time.

Support screen offering reassurance and next-step guidance without forcing action.

• App appears as a meditation or wellness app by default
• No visible recordings or suspicious UI
• Real functionality unlocks via long-press + PIN
• One-tap quick exit instantly returns to the decoy screen

• Victims document incidents using natural speech
• No typing or writing required
• Designed for trauma-informed, low-cognitive-load use

• Voice input is transcribed and analyzed automatically
• AI identifies what happened, when it happened, and key signals
• Users never need to think in legal terms

• AI asks gentle, contextual follow-up questions
• Helps recall details commonly lost under stress
• Converts fragmented memories into usable evidence

• Incidents are analyzed over time for escalation and repetition
• System adapts behavior when risk increases
• Designed to support prevention, not just documentation

• Chronological, private evidence log
• Stored securely and encrypted
• Exportable only when the user chooses to do so

• AI-powered guided meditation agent
• Built using ElevenLabs for calm, empathetic voice guidance
• Provides real value while reinforcing the decoy design

• ElevenLabs
  Used for empathetic voice interaction, incident documentation, and the guided meditation agent.

• Google Gemini
  Used for AI reasoning, narrative understanding, follow-up generation, evidence structuring, and pattern detection.

• Google Cloud
  Used for secure, encrypted processing and storage of sensitive data.

• Confluent
  Used for real-time event streaming to analyze incident patterns and detect escalation or crisis signals.

• Datadog
  Used for system observability, reliability monitoring, and risk signal tracking.

Each technology supports safety, reliability, and intelligence — not vanity features.

SafeVoice lowers the barrier to justice by:

• Prioritizing user safety above all else
• Preserving memory under trauma
• Turning lived experiences into usable evidence
• Helping victims prepare for legal or support systems without replacing them