Cernavo
Request beta access
Acceptable data policy

What you can upload

Private betaFor invited beta usersCernavo may reject any upload — this list is not exhaustive

Cernavo is a founder-hosted private beta. We accept low-sensitivity tabular data only. A simple test: if you would not be comfortable handing the file to a small startup operating on a single node, do not upload it.

Permitted by default

  • Tabular files in CSV, TSV, or Parquet format.
  • Low-sensitivity, sanitized, non-regulated business data — product catalogs, marketing exports, internal analytics extracts without customer PII, public-record datasets, and test fixtures.
  • Data you have the right to process and share with a third-party beta operator.

Default size limits

  • Default per-upload cap: 20 GB.
  • Absolute per-upload cap: 50 GB, by explicit approval only.
  • The server enforces the size limit. Your upload card shows the currently-enforced cap.

Not accepted by default

The categories below are prohibited by default. Cernavo may reject any upload that falls into one of these buckets, archive the associated conversation, and trigger the Delete now path. None of these are accepted without an explicit pre-approval conversation.

Protected health data
PHI / HIPAA-scope data, and anything that would identify a patient in a healthcare context.
Payment card data
PCI-scope data — full card numbers, magnetic-stripe data, CVVs, or anything subject to PCI DSS scope.
Regulated financial records
Account-level banking detail, brokerage statements, trading records, and anti-money-laundering case data.
Highly sensitive HR data
Compensation rosters tied to identified people, performance-review content, disciplinary records, immigration status, and medical-leave detail.
Legal privilege material
Attorney-client communications, litigation-hold data, and settlement discussions.
Government classified or controlled data
Any export-controlled, ITAR/EAR, CUI, or classified material.
Secrets and credentials
Passwords, API keys, access tokens, OAuth refresh tokens, private keys, signing keys, production database credentials, and internal service secrets.
Production security artifacts
Live security logs, IDS/SIEM extracts, vulnerability-scanner output tied to production hosts, and incident-response evidence.
Government IDs and identifying documents
Social-security numbers, passports, driver-license numbers, and national-ID numbers in identifiable form.
Highly identifying personal data
Outside the listed regulated buckets — biometric data, precise geolocation traces, private chat or message archives, and anything intended to identify or surveil specific individuals.
Format-restricted bundles
XLSX workbooks and ZIP archives are not supported in this release. Convert spreadsheet data to CSV, TSV, or Parquet before uploading.

When in doubt

  • Use a sanitized sample. Mask identifying columns. Use synthetic or test data when the goal is to evaluate the product.
  • Ask beta support at [email protected] before uploading anything you are unsure about — we would rather answer a question than reject a dataset after the fact.

The list above is not exhaustive

Cernavo may reject any dataset that is unsuitable for the private beta, even if it does not appear in the categories above.

If an upload is rejected

When a dataset is rejected, the bound conversation is archived and the run-scoped artifacts go through the Delete now path described on the data-handling page:

  • Run-scoped artifact references are tombstoned.
  • Application access is revoked — signed URLs and listings stop returning data.
  • Run-scoped key material is crypto-erased per the per-tenant boundary.
  • Physical garbage collection of object-store bytes runs asynchronously.

Audit metadata for the action itself is retained for operator accountability.

Last updated: June 2026