How Athena handles data
Athena is a credibility infrastructure platform. We document public claims from public sources. This page describes what data we collect, how we store it, who has access, and what we commit to.
Data Sources
Athena processes publicly available content only. We do not collect private messages, non-public posts, or data requiring user login to access.
Public videos, titles, descriptions, and captions via YouTube Data API v3.
Audio extracted and transcribed via Deepgram and GCP Speech-to-Text for public videos.
Historical price data sourced from Coinbase and Binance public APIs for outcome verification.
Structured claims extracted from transcripts via Google Gemini. No personal data is sent to LLM providers.
Retention Policy
Claims, scores, and dispute records are retained indefinitely to preserve the integrity of the credibility ledger. Score history is append-only and is never deleted or modified.
| Data type | Retention |
|---|---|
| YouTube video metadata | Indefinite |
| Transcripts | Indefinite |
| Extracted claims | Indefinite (append-only) |
| Credibility scores | Indefinite (append-only history) |
| API keys | Until revoked by user or admin |
| Dispute records | Indefinite |
| Email preferences | Until user deletes account |
API Key Handling
API keys are hashed before storage. We never store plaintext keys after initial issuance.
- Keys are scoped to read-only access by default.
- Keys can be revoked at any time from your account settings.
- Failed authentication attempts are rate-limited.
- Key metadata (creation date, last used) is logged for audit purposes.
- We do not share API keys with third parties.
Subprocessors
Athena uses the following third-party services in its data processing pipeline. Listed for transparency — not indicative of partnership or endorsement.
| Subprocessor | Role |
|---|---|
| Vercel | Frontend hosting and edge delivery |
| Google Cloud Run | Backend API and microservices |
| Google Cloud SQL | Primary database (PostgreSQL) |
| Deepgram | Speech-to-text transcription |
| Google Gemini | Claim extraction AI |
| Coinbase / Binance | Market price data for outcome verification |
| Webshare | Proxy infrastructure for data collection |
| Redis / Celery | Task queue for async pipeline jobs |
Independence Commitments
Dispute & Correction SLA
Any subject, viewer, or researcher can file a dispute if they believe a claim was extracted incorrectly, an outcome was verified against wrong data, or a factor was computed from bad inputs.
Security Contact
For security disclosures, data questions, or deletion requests, contact us at:
We follow industry-standard security practices for a cloud-hosted SaaS platform. We are not yet SOC 2 certified.