A safe upload decision starts with understanding storage, decryption access, training and review policies, logged metadata, and deletion controls.
Review every extracted value against the original document before saving the imported record.
A safe upload decision starts with understanding storage, decryption access, training and review policies, logged metadata, and deletion controls.
Before you begin
Original source file
Have the original report or genetics file ready so you can compare every imported value with it.
File quality
Unlock password-protected files and obtain a complete, readable original before importing.
Methodology
The workflow was documented against the current Catina import flow and the linked provider export instructions where applicable.
Read the full scoring, sourcing, and corrections policyWhat you'll need
- Open the provider's current privacy and AI-training documentation.
- Prepare a copy of the report and remove unnecessary identifiers where practical.
- Keep the original report available so you can verify every value in the AI output.
Steps
- 1
Identify the product surface
A dedicated health space may use different controls from a regular consumer chat.
- 2
Map storage and inference separately
Ask where the file rests, where plaintext processing occurs, and which partners participate.
- 3
Check training, review, and retention
Read defaults as well as optional controls and safety-review exceptions.
- 4
Minimize the upload
Remove names, addresses, account numbers, and pages outside the analysis.
- 5
Verify the output
Compare every value, unit, range, and date with the source report.
What crosses the network boundary
The selected source file is read in the browser. When an import needs sensitive AI extraction or explanation, the request body is encrypted in the browser for an attested Tinfoil enclave. Decryption happens exclusively inside the verified enclave. Catina's relay handles authentication, routing, limits, and billing metadata.
What is stored locally
After you confirm the extracted values, the supported record is saved in the encrypted local Health Vault. Standard vault unlock is backend-assisted. Extra Local Protection separately adds a passphrase-derived local encryption layer.
Review and confirmation
Compare names, dates, units, reference ranges, flags, and values with the original. Correct or remove any mismatch before saving.
Limitations
- Upload requirements vary across legal, employment, insurance, and clinical contexts.
- Evaluate both transport encryption and decryption access at the processing destination.
Troubleshooting
- If text extraction fails, use the original provider-issued PDF for the clearest input.
- If units or reference ranges look wrong, edit the row before saving and keep the original file for comparison.
- If enclave verification stops the request, try again later or contact support.
Verification Center evidence
Open Catina's Verification Center before a sensitive AI action to inspect the current enclave verification status and implementation evidence. A verification failure stops the sensitive request before transmission.
FAQ
What happens to my file when I use an AI feature?
The saved copy stays in your local Health Vault. The data needed for an AI request is encrypted on your device and opened exclusively inside a verified secure enclave for processing. Encrypted content passes through Catina and the supporting infrastructure, and the enclave discards the protected health content after processing. The service handles limited technical information needed to authenticate your account and route the request.
How should I use Catina's results?
Use Catina for informational health insights. Confirm abnormal or concerning results with a qualified healthcare professional.
Sources
- Catina privacy architectureCatina · verified August 20, 2026Supports: Catina storage, Catina AI processing, Catina limitations
- Authoritative Catina product informationCatina · verified August 20, 2026Supports: Catina features, Catina availability, Catina platform
- Catina plans and pricingCatina · verified August 20, 2026Supports: Catina pricing, Catina plan limits
- Proxy encrypted requests through your backendTinfoil · verified August 20, 2026Supports: Encrypted request bodies, Proxy-visible headers, Client verification
- Backend infrastructure and attestationTinfoil · verified August 20, 2026Supports: Secure enclave attestation, Client-side verification
- Health in ChatGPTOpenAI · verified August 20, 2026Supports: Health in ChatGPT availability, connectors, training controls, medical boundary
- How your data is used to improve model performanceOpenAI · verified August 20, 2026Supports: ChatGPT consumer training controls, Temporary Chat
- Gemini Apps Privacy HubGoogle · verified August 20, 2026Supports: Gemini data controls, retention, human review
- How do you use personal data in model training?Anthropic · verified August 20, 2026Supports: Claude consumer training controls, safety review
- What is Perplexity Health?Perplexity · verified August 20, 2026Supports: Perplexity Health availability, connectors, dashboard, retention
- Privacy policyVenice AI · verified August 20, 2026Supports: Venice provider handling, zero-retention agreements
Build your private health record
Import lab reports from any lab or country, add supported raw genetics and lifestyle context, track changes over time, and generate personalized insights. Confirm extracted values before saving them to your encrypted local Health Vault.