Standards
Compliance is system design. Ethical data practices are not optional — they are foundational to data that can be used for research, commercial deployment, and long-term AI development.
All data involving human participants is generated with explicit, informed consent.
For speech and voice data, we work with over 200 professional performers — actors, voice actors, announcers, and narrators — under consent frameworks that explicitly define usage scope, region, duration, derivative rights, and withdrawal conditions.
Informed Consent: Participants understand how their data will be used before recording
Usage Scope: Clear documentation of permitted uses, regions, and durations
Derivative Rights: Explicit terms for derivatives, retraining, and future applications
Withdrawal Rights: Participants can withdraw consent according to agreed terms
All data is maintained with sample-level traceability to consent conditions, withdrawal terms, and derivative use permissions. The system is designed so that unauthorized use is technically prevented.
Consent is not a checkbox. It is a design requirement.
Data generated for different purposes is never mixed without explicit design.
Research data and commercial data are separated at generation
Sample-level traceability to consent conditions
Audit trails for compliance verification
Clear documentation of data provenance
This enables compliance with EU research requirements, international commercial use, and future regulatory frameworks.
We design data with awareness of how it will be used in AI systems.
Bias Awareness: Demographic representation is designed, not incidental
Harm Prevention: Data is not generated for applications designed to harm
Transparency: Data characteristics are documented, not hidden
Accountability: We maintain records enabling post-hoc evaluation
When working with professional performers and speakers:
Fair compensation for data contribution
Clear terms for voice and likeness usage
Protection against unauthorized derivatives
Ongoing relationship for re-recording and expansion
Professional relationships enable reproducibility. Short-term extraction prevents it.