1. Overview
Bai – Bisaya AI ("we," "our," or "us") is committed to protecting your privacy. This Privacy Policy explains what personal data we collect, why we collect it, how it is processed across our technical infrastructure and AI backends, and your rights under the Philippine Data Privacy Act of 2012 (Republic Act No. 10173) and other applicable privacy regulations.
2. Age Restriction (16+)
Bai is intended solely for individuals aged 16 and older. We do not knowingly collect personal data from anyone under the age of 16. If we become aware that an account belongs to a person under 16, we will promptly terminate the account and delete all associated records.
3. Information We Collect
Account and Authentication Data: We collect your verified email address and basic public profile details (e.g., name or avatar image if provided by Google) when you authenticate via Google Sign-In or request an OTP Magic Link through Better Auth.
Conversations, Prompts, Voice Inputs, and Feedback: We collect conversation records across all interactions—including authenticated users and guest sessions—comprising text search queries, chat prompts, voice audio clips (for multimodal speech recognition), AI-generated responses, and human-in-the-loop feedback/corrections. These records are gathered specifically to build, align, and fine-tune native Bisaya AI foundation and speech models that listen, speak, and reason natively in Bisaya. All records undergo automated PII sanitization (masking Philippine mobile and landline numbers, emails, and financial identifiers) before being ingested into training corpora.
Generated Static Content: When you create shareable content cards, the application renders them as HTML components and exports them as PNG images directly within your client browser. These elements are preserved solely as part of your saved conversation thread in the database, rather than in separate public object storage buckets.
Transaction and Credit Records: We maintain a record of your available "load" credit balance, consumption history, and transaction references provided by Polar.sh. We do not collect or store full credit card numbers or banking credentials.
Network and Technical Metadata: Our hosting provider (Cloudflare Workers) automatically processes technical metadata such as your IP address, browser user-agent, device characteristics, and request timestamps to maintain edge security, caching, and rate limiting.
4. Third-Party Service Providers and Data Processors
We engage trusted third-party services to deliver Bai. Data shared with these processors is limited to what is strictly necessary:
- Hosting and Edge Infrastructure: Cloudflare Workers hosts the web application and handles edge routing and request security. Cloudflare R2 provides encrypted object storage for 90-day voice processing and training dataset archives.
- Authentication: Better Auth handles session management, token issuance, and OTP distribution. Google OAuth verifies credentials when you select Google Sign-In.
- Payment Processing: Polar.sh manages the checkout and billing processes for load credit packages.
- AI Model Inference Providers: User prompts and conversation context are
transmitted to backend AI APIs to generate responses:
- Gemma 4 via Cloudflare Workers AI: Cloudflare processes inputs and outputs for inference without using customer content to train AI models or improve services.
- Gemini 3.5-Flash & Flash-Lite via Google Gemini API: Processed according to Google's API service terms. (When operating under billing-enabled enterprise/paid API accounts, Google does not use prompts to train models or conduct human review).
- GPT OSS 120B via Groq: Groq does not retain inference prompts by default, does not use customer prompts to train foundation AI models, and only retains temporary logs for up to 30 days for service reliability and abuse prevention.
5. Cookies and Local Session Storage
Bai utilizes strictly functional cookies and session tokens managed by Better Auth to maintain authenticated user sessions and verify account states. We do not use persistent advertising cookies or sell data to third-party ad networks.
6. Data Retention, Model Training, and Account Deletion
Retention Period for Text Conversations: We retain conversation history for active accounts to provide continuous session continuity and de-identified training datasets for native Bisaya language foundation and speech model development.
Voice Audio Retention (90 Days): Raw voice recordings (.webm audio files) stored in Cloudflare R2 are retained for a maximum of ninety (90) days to allow for acoustic feature extraction, ASR/TTS training ingestion, and safety moderation, after which raw audio files are automatically purged. De-identified textual transcriptions and phonetic metadata remain part of the training dataset.
7. Security Measures
We implement administrative and technical safeguards—including TLS encryption in transit, Cloudflare edge security defenses, and secure session tokens—to protect your personal information against unauthorized access, loss, or alteration.
8. Changes to This Privacy Policy
We may update this Privacy Policy to reflect changes in our operational procedures or applicable legal frameworks. When revisions are made, the updated Effective Date at the top of the policy will be modified.
9. Contact Us
For any privacy questions, data access requests, or deletion inquiries, reach out to contact@kumustabai.com.