Alpha core in development
Sensitive context protection before AI tools receive it.
Orisan Guard is being built to detect sensitive context, rewrite it safely before it reaches AI tools, and produce evidence-safe events without uploading raw prompts.
Core question
Can this prompt or context be safely sent to an AI tool?
Guard is not released as a browser extension yet. The current work is the local alpha core: classification, placeholder planning, redaction, and evidence-safe benchmarking.
Future flow
Raw prompt to safe rewrite to evidence event.
01
Raw prompt
A user drafts a prompt or pastes context intended for an AI tool.
02
Sensitive context detected
Guard classifies spans such as credentials, internal hosts, identifiers, and private material.
03
Safe rewrite
Sensitive spans are replaced with collision-resistant placeholders while preserving syntax where possible.
04
Evidence event
The output records what was protected without storing raw prompt payloads.
Alpha core
The foundation is local and evidence-safe first.
Honest status
Guard is under active development, not a finished platform.
Product boundary