Company

Transparency

What we disclose, why we disclose it, and where to find the details.

Our transparency commitment

Transparency is foundational to trust in a forensic analysis platform. If you can't understand how our assessments are produced, you can't meaningfully evaluate them. We publish information about our methods, accuracy, limitations, and practices so that users can make informed judgments.

What We Publish

Transparency disclosures

Methodology

Complete documentation of our scoring methodology including module weights, confidence calibration, evidence fusion process, and verdict thresholds.

Read methodology

Detection limitations

Honest disclosure of scenarios where our analysis is less reliable: novel generators, adversarial attacks, heavy compression, short clips, and more.

Read limitations

Validation results

Internal validation study with accuracy metrics, per-module performance, false positive rates, and known failure modes.

Read validation study

Ethics framework

Our principles for responsible AI detection: honest communication, false positive awareness, non-weaponization, privacy protection, and equitable access.

Read ethics

Module documentation

Detailed documentation of all 15 forensic modules including what they analyze, how they work, and what they can and cannot detect.

Read module docs

Research publications

Educational content about deepfake detection, video forensics, and media authenticity. We publish what we know to help the broader community.

Read research

Data handling

Video storage: Uploaded videos are processed and stored securely in your account. We do not share your videos with third parties.

Analysis results: Forensic reports are associated with your account and visible only to you. You can delete them at any time.

Training data: We do not use your uploaded videos to train our models without explicit, opt-in consent.

Analytics: We collect anonymized usage analytics (page views, feature usage) to improve the platform. No video content is included in analytics.