Contributor Data & AI License
Version: September 26, 2026 · UMI AI, Inc.
Commercial AI and dataset use
You authorize SignRush to use and license your submitted recordings, interpretations, corrections, quality ratings, labels, translations, annotations, metadata and derived data for commercial AI training, evaluation, benchmarking and dataset products, including licensing those materials to third parties. This includes retained failed or rejected submissions for quality analysis and research with their actual validation status; rejection never makes an incorrect label correct.
Your license grant
You retain ownership of your original contribution. You grant UMI AI, Inc., operating SignRush, a worldwide, non-exclusive, perpetual, transferable, sublicensable, royalty-free license, irrevocable to the extent permitted by law, to capture, store, reproduce, process, analyze, adapt and commercially use these contributions for the purposes described here. Rights are subject to applicable privacy law and legal restrictions. This is not a claim of ownership over ASL or over your identity.
Recordings, likeness and derived features
The license includes your image, face, hands, body movement, signing, performance, likeness and audio if present in a submission. The current recorder requests silent video, not microphone access. Face and hand features, hand geometry or motion, body pose, movement patterns, keypoints, landmarks and embeddings may be extracted for AI/data purposes. Such features can be identifying or sensitive. No permission for identity authentication or unrestricted public promotion is granted by this agreement.
Models and transformations
Permitted purposes include model development, training, fine-tuning, validation, safety testing, inference research, evaluation and benchmarking. Processing may include cropping, resizing, compression, normalization, frame extraction, segmentation, transcription, translation, annotation, feature extraction, aggregation, de-identification where appropriate, combining with other data, synthetic or derived training data, model outputs, weights and other learned representations. Trained models may retain learned statistical information after an original recording is no longer actively used.
Licensing and recipients
SignRush may package raw recordings and derived data in commercial datasets; license or sublicense private access to third-party AI developers, research organizations and other dataset customers; provide API, subscription, training or benchmark access; and transfer the granted rights in connection with a sale, merger, acquisition or financing. Authorized reviewers and processing providers may access material for validation and operation. Public marketing use is excluded below.
Review contributions and provenance
Your independent interpretations, corrections, quality ratings and other annotations become licensed corpus contributions. SignRush preserves original answers and provenance, including prompt/version, source hashes, review history, consent receipts and signer-level training/evaluation designations. A prompt is intended meaning, not a verified transcript. Disputed material remains restricted from training until resolved and separately eligible.
Compensation and ownership
Task rewards are governed by the task terms and Terms of Service. No ongoing royalty, residual or share of revenue is due unless separately agreed in writing. You do not receive ownership of resulting models, curated datasets or products. Game points do not automatically create a cash balance.
Stopping, deletion and surviving rights
You may stop future participation through Account and request access, deletion or restriction as described in the Privacy Notice. Applicable privacy-law rights will be honored. Stop contributing blocks new participation and future eligible exports in the current system while preserving your original agreement and license records. Stopping does not itself cancel the license for existing contributions, subject to applicable law. It does not automatically undo completed lawful licenses or remove learned model information. Lawful audit records, backups, aggregate or de-identified information and learned representations may remain where permitted; deletion or restriction obligations take precedence where required.
No public promotional permission
This agreement does not authorize identifiable recordings in advertising, social media, public website demos or investor presentations. Such display requires separate permission. The current onboarding does not request or grant it. Aggregate research findings and benchmark results may be published without publicly identifying contributors.