AI Use Statement
Last updated: April 2026
This AI Use Statement explains where Stanza uses AI-assisted systems, what those systems do, and what they do not do.
Defined Terms
- AI-assisted system — A software component that uses machine-learned models to generate or classify outputs.
- Automated moderation signal — A model-generated flag or score used to prioritize review, not a final enforcement decision by itself.
- Speech-to-text (STT) — Automated conversion of spoken audio into text transcripts.
Where We Currently Use AI
- Spoken-word transcription — When creators enable STT on audio submissions, transcripts can be generated automatically.
- Transcript safety checks — Automated moderation models may analyze transcript text for policy-risk categories and create moderation signals.
- Operational automation — We may use automated ranking and processing systems for feed/search quality and abuse prevention.
What We Do Not Claim
- We do not claim AI outputs are always accurate. Transcripts and moderation classifications can contain errors.
- We do not use AI as the sole source of truth for legal, safety, or account-level enforcement decisions.
- We do not guarantee equal model performance across all accents, dialects, languages, or recording conditions.
Human Oversight
AI-generated moderation signals are intended to support reviewers and internal tooling. Final enforcement actions may include human review based on severity, confidence, and policy context.
User Controls
- Creators can enable STT for eligible spoken-word submissions.
- Creators can edit transcript text where editing is available in the product flow.
- Users can report incorrect, harmful, or policy-violating outputs through in-app reporting or support channels.
Data Handling
- Audio and transcript data are processed to provide requested product features (for example transcription and safety checks).
- Model providers and infrastructure services may process data on our behalf under contractual and technical controls.
- Retention and deletion of account/content data follow our Privacy Policy and product retention rules.
Known Limitations
- Background noise, overlapping voices, low bitrate, or non-standard pronunciation can reduce transcript quality.
- Automated policy classification may produce false positives or false negatives and should be interpreted as a signal, not a verdict.
Updates to This Statement
As we add or change AI-assisted features, we will update this statement to reflect material changes in AI use, controls, or oversight.
Contact
Questions about AI use on Stanza can be sent to legal@stanza.ink.