Updates
Product updates and beta progress
See what's shipping, what's improving, and what Privian is actively working on during beta.
Built transparently. No fixed dates, no promises — just a grounded view of what's moving.
- BYOK provider model
- Zero raw prompt retention
- Documented masking types
- Transparent beta limitations
- Public architecture & security
Latest
Latest shipped improvements
Recent changes to the Privian platform and the public site.
- Shipped
Solution pages expanded
Customer support, internal copilots, AI features for SaaS and prototyping pages rebuilt with real architecture, examples and beta limitations.
- Shipped
Product pages expanded
LLM Gateway, PII Masking, Prompt Security and Rehydration pages now describe the actual contract, supported entities and zero-retention flow.
- Shipped
Marketing and docs consistency pass
Aligned all public copy with the real gateway: small JSON contract ({model, prompt, stream}), supported entity set, and current beta scope.
- Shipped
Technical documentation expanded
Added grounded concepts, getting-started, guides and API reference based on the actual gateway implementation.
- Shipped
SEO and AI-search foundations
Added glossary, category pages, breadcrumbs and structured data so the site is readable to search engines and LLM-based search.
In progress
Currently improving
Work we're actively researching, planning or building. No dates — priorities can shift based on feedback.
- PlannedHigh priority
Norwegian fødselsnummer masking
Investigating dedicated detection and validation for Norwegian national identification numbers.
- PlannedHigh priority
Custom entity masking
Exploring user-defined sensitive entities so teams can mask domain-specific identifiers.
- PlannedMedium priority
Native provider token streaming
Working toward forwarding real provider token streams instead of the current artificial chunking.
- PlannedMedium priority
OpenAI messages[] compatibility
Investigating an optional input shape compatible with OpenAI-style messages arrays.
- ResearchingMedium priority
Tool / function calling support
Researching how to expose provider tool-calling through the gateway without weakening privacy guarantees.
- PlannedLow priority
Expanded developer ergonomics
Improving error shapes, request introspection and dashboard feedback for integrators.
Direction
Exploring next
Areas we are actively exploring during beta. Priorities may evolve based on feedback and product learning.
Custom sensitive entity definitions
Let teams define their own entity types and matchers.
Expanded masking coverage
More built-in entities, locales and validators over time.
Improved developer ergonomics
Better error messages, examples and per-key debugging surfaces.
Streaming improvements
Move from artificial chunking to native provider streams.
Prompt workflow flexibility
Explore additional input shapes without breaking the privacy-first contract.
Transparency
Known beta limitations
What Privian does not do today. Surfacing this is part of how we build trust.
No native provider token streaming
The gateway currently returns responses in artificial chunks rather than forwarding real provider token streams.
No tool / function calling
Provider tool-calling and function-calling features are not exposed through the gateway yet.
No OpenAI SDK drop-in
Privian uses a lightweight JSON contract ({model, prompt, stream}) rather than mirroring the OpenAI SDK surface.
No messages[] format yet
Requests use a single prompt string today. Multi-message inputs are on the exploration list.
Bounded entity coverage
Detection covers a defined set of personal identifiers, financial / national IDs and developer secrets — not arbitrary content.
No Norwegian fødselsnummer support yet
Dedicated detection for Norwegian national IDs is planned but not currently shipping.
No custom user-defined masking yet
Custom entity definitions are not configurable in the current beta.
No HIPAA / SOC 2 / PCI claims
Privian is in beta and does not currently claim formal compliance attestations.
Feedback
Have feedback or feature ideas?
We're actively improving Privian during beta and welcome practical feedback from developers and teams building with AI.
Lightweight by design. A richer feedback and voting system may come later — for now, email reaches the team directly.
FAQ
Frequently asked questions
- Is this a public roadmap with delivery dates?
- No. The updates page is a transparency log. It surfaces shipped work, current direction and known beta limitations without committing to dates.
- How often is this page updated?
- Whenever meaningful work ships or priorities shift. Items here are grounded in the actual implementation, not marketing plans.
- Can I influence what gets built?
- Yes. Practical feedback from developers and teams building on Privian directly shapes priorities during beta.
- Why are there beta limitations listed publicly?
- Surfacing limitations is part of building trust. It lets developers evaluate Privian honestly before integrating.
Start building
Build safer AI while Privian evolves
Follow product progress and start building with privacy-first LLM workflows.