DocsPrivacy & Data
Reference

Privacy & Data

How your data is stored, who can see it, and what Tyrex never does with it — including the fine-tuning service's fire-and-forget policy.

Privacy is not a feature tier at Tyrex AI — it's the architecture. Your AI is private by default, and the platform is designed so your data is never shared, sold, or used to train a shared model.

What Tyrex never does

  • Never trains on your data. Your files, chats, and personas are never used to improve a shared model.
  • Never shares your AI with others. Your persona and its knowledge base are scoped to your account.
  • Never reads your data for marketing. No analytics on your document contents, no profiling.

How your data is stored

  • Knowledge base — uploaded files are stored encrypted and indexed in a private vector database scoped to your AI (persona). Retrieval only ever happens for your AI, in your sessions.
  • Chats & sessions — conversation history is stored per account, tied to the persona you were talking to.
  • Memories — durable facts are explicit and editable; delete any memory at any time.
  • Storage quota — your knowledge counts against your plan's storage (100 MB free, 10 GB Pro, 100 GB Elite).

Scoping: per-AI, not per-account-wide

Knowledge is attached per persona. Your teacher AI cannot access your business AI's files — even within the same account. This isolation is enforced at the retrieval layer, not just in the UI.

Marketplace & PII

When you sell a dataset, personally identifiable information is stripped automatically before the listing can be published — names, phone numbers, addresses, and similar fields are removed. Paid listings are also manually reviewed before going live.

The fine-tuning service: fire-and-forget

Tyrex's optional fine-tuning service (clearly labeled, paid, separate from the free loop) follows a strict data policy:

  • Training data you upload goes directly to an ephemeral GPU environment for the job.
  • It is purged when the job finishes — not stored, not logged, not added to any database, and never used as Tyrex training data.
  • The only thing that "remembers" is the resulting model file, which you download.
  • Result files are kept for a limited retention window (until you download, plus a grace period), then purged.

In short: for fine-tuning, your data is used for the job and nothing else.

  • Tyrex AI operates in compliance with Indonesian personal data protection law (UU PDP).
  • Your full legal terms and privacy policy: Privacy Policy · Terms of Service.

Model transparency

Tyrex AI runs established open models as its engines. The platform is transparent about this — the public model names (RAPTOR, CARNO, REX) are Tyrex's product brands, and details of the underlying technology are documented in our privacy materials rather than hidden.

Next: back to Getting Started, or browse the Docs home.