Deepfake Detection Needs Better Data
Current detection models are limited by small, narrow datasets that lack detailed annotations. Without large-scale, well-annotated data covering the full range of deepfake techniques, detection tools struggle to keep pace.
500,000+
Daily deepfakes created
96%
Go undetected
Limited
Labeled datasets
Three Steps to Better Data
Submit Deepfake Media
Upload images, videos, or audio with context about the source and deepfake technique used.
Annotate Submissions
Classify the deepfake type, rate quality, and flag notable characteristics of submitted media.
Peer Review Annotations
Validate annotation accuracy through consensus-based review to ensure data quality.
Earn Recognition for Your Contributions
Every contribution earns points that unlock new levels and recognition within the community.
Built for Rigorous Research
Every design decision prioritizes data quality, reproducibility, and academic rigor.
Peer Review
Multiple reviewers validate every annotation
Structured Taxonomy
Standardized deepfake categories and schemas
Quality Scoring
Consensus-based annotation quality assurance
Open Dataset
Open access for building detection models