Skip to main content
Back to Home

Trust & Safety Policy

Last updated: December 28, 2025

Our Commitment to Responsible AI

We understand the concerns regarding Generative AI risks. While we utilize industry-leading providers (ElevenLabs, Google) who have their own safety layers, SceneFlow AI enforces its own strict application-level guardrails to prevent misuse before it ever reaches our upstream providers.

SceneFlow AI operates as an active gatekeeper—not a pass-through service. Our Compliance Layer sits between users and AI providers, implementing multiple safeguards to ensure responsible use of generative AI technology.

The Four Pillars of Our Guardrails

A. Identity Verification ("Voice Captcha")

CRITICAL FEATURE

Mechanism

We use ElevenLabs' Voice Verification technology. When creating a voice clone, the user must record themselves speaking a specific, randomized prompt (e.g., "I, [User Name], consent to having my voice cloned by SceneFlow").

Guardrail

If the voice in the consent prompt doesn't match the voice in the uploaded sample, the clone request is automatically rejected. Users cannot impersonate others without physical access to the person.

Goal

Prove users can only clone their own voices or voices they have explicit rights to use.

Trust Statement: "We enforce biometric verification on all voice clones. Users cannot impersonate others without physical access to the person providing live consent."

B. Content Moderation (Input Scanning)

Mechanism

All text and media content is processed through our multi-layer moderation system before being sent to AI providers. This includes:

  • OpenAI Moderation API for text content screening
  • Custom keyword filtering for fraud and scam detection
  • AI-powered visual moderation for generated images and video (via Hive AI)
  • Audio content analysis for voice-generated outputs

Guardrail

High-risk content categories are blocked specifically for Custom/Cloned voices:

  • Hate speech and discriminatory content
  • Fraud-related terminology and scam scripts
  • Political disinformation and election interference content
  • Impersonation of public figures
  • Content depicting or targeting minors

Standard ("Stock") voices have more lenient restrictions as they present lower impersonation risk.

Goal

Formalize content review to prevent deepfake abuse and misuse of synthesized voices.

Trust Statement: "We pre-scan all scripts assigned to custom voices to prevent deepfake abuse. Prohibited content is blocked before it ever reaches our AI providers."

C. Tiered Access ("Trust Score")

Access Tiers

Tier 1: New User

Mechanism: Can only use our curated "Golden Set" of Stock Voices from ElevenLabs.

Guardrail: High-risk features (voice cloning) are restricted for new accounts, eliminating quick "burner account" abuse.

Tier 2: Verified/Paid User

Mechanism: Can access Voice Cloning after a minimum 7-day account age or after manual review of their project by our Trust & Safety team.

Guardrail: Payment verification and account history review before enabling cloning capabilities.

Trust Statement: "High-risk features are behind a 'Trust Wall.' We manually review account standing before enabling voice cloning, eliminating 'burner account' abuse."

D. Watermarking & Attribution

Mechanism

All generated audio includes forensic watermarking through ElevenLabs' Audio Native technology. Video content utilizes Google's SynthID invisible watermarking for AI-generated media.

Guardrail

If a piece of audio or video content leaks or is used maliciously, it can be traced back to:

  • The SceneFlow platform
  • The specific user account
  • The exact generation timestamp

Goal

Enable rapid identification and banning of bad actors while supporting law enforcement investigations when required.

Trust Statement: "All generated audio and video carries forensic watermarks, allowing us to trace and ban bad actors instantly."

Liability & Compliance

Liability Shielding

Our Terms of Service explicitly transfer liability for generated content to the user. Users agree that they are solely responsible for:

  • The content of prompts and scripts submitted to the platform
  • How generated content is used, distributed, or published
  • Ensuring they have rights to clone any voices they upload
  • Compliance with applicable laws in their jurisdiction

Audit Logging

We maintain comprehensive audit logs of every generation, including:

  • User ID: Account identifier linked to the generation
  • Prompt/Script: The input text used for generation
  • Audio Fingerprint: Unique identifier for the generated content
  • Timestamp: Exact date and time of generation
  • Voice Used: Whether stock or cloned, with voice identifier

These logs are retained to assist law enforcement if requested, support content attribution investigations, and enable rapid response to abuse reports.

Reporting Abuse

If you encounter content generated by SceneFlow AI that you believe violates our policies or is being used for malicious purposes, please report it immediately:

  • Email: abuse@sceneflowai.com
  • Include: URLs, screenshots, or audio/video files if available
  • Response Time: We investigate all reports within 24 hours

Summary: SceneFlow AI Guardrails

Non-Consensual Cloning Block: We mandate "Voice Captcha" verification. Users cannot clone a voice without a live recording of that person giving consent.

Input Moderation: We use an intermediate AI layer to scan scripts for prohibited content (fraud, hate speech) before audio generation is permitted.

Tiered Access: Voice Cloning is a restricted feature available only to verified, paid accounts—eliminating "burner account" abuse.

Forensic Watermarking: All generated content carries invisible watermarks for traceability and attribution.

Liability Shielding: Our ToS explicitly transfers liability for generated content to the user, and we maintain audit logs (User ID + Prompt + Audio Fingerprint) to assist law enforcement if requested.

Contact Our Trust & Safety Team

Trust & Safety: trust@sceneflowai.com
Abuse Reports: abuse@sceneflowai.com
Legal Inquiries: legal@sceneflowai.com
Website: https://sceneflowai.studio

SceneFlow AI - AI-Powered Video Creation