Non-transfer by design
Original files and databases remain in the enterprise environment.
ENTERPRISE AI DATA OPERATING SYSTEM
Scale AI operations
without surrendering
data control.
NEWELYS turns data already inside the enterprise into governed AI assets—connecting productization, rights, diligence, model evaluation, settlement and revocation in one operating system.
ORIGINAL DATA STAYS IN PLACE
POLICY-BOUND AI ACCESS
AUDITABLE BY DESIGN
POLICY-BOUND AI ACCESS
AUDITABLE BY DESIGN
- 01CONNECT
- 02MANIFEST
- 03POLICY
- 04ROOM
- 05METER
- 06LIFECYCLE
33architecture modules
13Foundation systems
20Execution controls
8Industry models
2022Founded in Seoul
WHY NEWELYS
The enterprise data problem is not collection.
It is controlled use.
Enterprise data already exists. The challenge is making AI execution environments understand and enforce who may use it, for what purpose, within what scope, and for how long.
PLATFORM ARCHITECTURE
Keep originals inside.
Enable AI through governed access.
Installed connectors map location, hash, permission, and state metadata, while the policy engine governs productization, delivery, and audit as one continuous flow.
Purpose-based rights
Reading, RAG, evaluation, training, download, and export rights are enforced separately.
Verifiable operations
Usage, changes, revocation, settlement, and reapproval remain auditable.
PLATFORM SUITE
Six product layers operate the full enterprise data lifecycle.
Not a collection of point tools, but a control infrastructure between existing systems and AI runtimes.
NEWELYS Connect
Install local connectors across DB, DMS, CMS, file servers and APIs to connect only selected data.
- Local connector
- Source map
- On-prem control
NEWELYS Manifest
Package scope, provenance, rights, quality, de-identification, pricing and state into a machine-readable product object.
- Manifest
- Trust vector
- Product card
NEWELYS Policy
Enforce RAG-only, eval-only, no-train and cleanroom-only conditions through tokens and runtime brokers.
- Policy vector
- Access token
- Runtime broker
NEWELYS Room
Provide buyer diligence rooms and scorer-only model evaluation rooms without exposing source data.
- Buyer room
- Safe sample
- Cleanroom eval
NEWELYS Meter
Measure RAG, API and evaluation use plus evidence contribution to create settlement candidates and audit proof.
- Usage event
- Attribution
- Settlement proof
NEWELYS Lifecycle
Synchronize rights changes and revocation across RAG indexes, APIs, caches, access grants and settlement states.
- Revocation
- Reapproval
- Propagation proof
GOVERNED DELIVERY
The same data is delivered differently for each purpose.
Execution rights such as RAG-only, Eval-only, No-train, and No-download are enforced and revalidated at query time by user, organization, country, and time. Switch the purpose on the right to see how the same request resolves.
Explore the six core technologies ↗Requesteranalyst@corp.example
Data productinternal-docs · v3
ConditionsKR · on-prem · 2026
Purpose of use
- Rights-holder consent—
- De-identification—
- Usage scope policy—
- Audit trail binding—
···
ORIGINAL DATA · NOT MOVED — CONNECTED BY REFERENCE AND POLICY ONLY
ENTERPRISE PROGRAMS
Build a data operating model, not just another technology stack.
NEWELYS connects data, security, legal and AI teams from assessment and design through pilot and operational transition.
Enterprise Data Productization
Build enterprise data inventory, rights and de-identification checks, manifests, catalogs and buyer rooms.
↘Governed RAG & Copilot
Implement enterprise RAG with query-time validation of department, role, project and purpose-based rights.
↘AI Evaluation & Cleanroom
Design scorer-only evaluation systems that validate external models and APIs without exporting source data or labels.
↘Data Rights & Revenue Operations
Connect access tokens, usage snapshots, attribution, provider settlement and revocation into an operating process.
↘CORE TECHNOLOGY
Six technology domains protect the full data lifecycle.
INDUSTRY SYSTEMS
Industry-native data structures on one common control plane.
Combine industry-specific source units, ground truth structures, rights models and evaluation rules on a common platform.
01 / PUBLIC AI
Public AI
Evidence, counterexample, refusal and uncertainty evaluation for public AI
02 / CONTENT & EDUCATION
Content & Education
Rights-bound data products protecting source content and answer keys
03 / INSURANCE AI
Insurance AI
OCR ground truth, damage labels and scorer-only evaluation
04 / LEGAL & HR
Legal & HR
Anonymous case data and evidence-based RAG evaluation
05 / ENTERPRISE KNOWLEDGE
Enterprise Knowledge
Enterprise RAG revalidating department, role and project rights
06 / HEALTHCARE & IMAGING
Healthcare & Imaging
Image data rooms combining consent, de-identification and cleanroom controls
07 / FRANCHISE INTELLIGENCE
Franchise Intelligence
Turn POS, VOC and local-market data into permissioned analytical assets
08 / CONSTRUCTION & CLAIMS
Construction & Claims
Structure defect images, work orders and estimates into evaluation data
Deployment and operations designed for enterprise scale.
On-premise / VPC
Deployment that keeps source and sensitive data inside organizational security boundaries.
Private SaaS
Operate dedicated tenants, policy control planes and audit/settlement modules.
Role-based Operations
Role-specific operations for data, legal, security, privacy, AI and procurement teams.
Policy-bound Runtime
Bind policies directly into RAG, evaluation, API and cleanroom execution paths.
ARCHITECTURE
33 architecture modules form one operating system.
Thirteen foundation systems and twenty execution controls create continuous technical boundaries from connection through revocation.
33architecture modules
13foundation systems
20execution controls
1AI data operating system
Installable Connector for Publishing Selected Enterprise Data as AI Data Products
Connects local enterprise sources to a central control plane so selected data can be productized without replacing existing systems or moving source files.
Non-transfer AI Data Product Manifest
Packages source references, scope, quality, de-identification, rights, price and state into a machine-readable product object while the original data stays in place.
Purpose-based AI Data Usage Policy and Access Enforcement
Separates RAG, evaluation, training, fine-tuning, cleanroom, API, download and export rights, then enforces them inside the actual runtime.
Pre-purchase Review Room and Non-disclosing Due Diligence Package
Lets buyers review metadata, quality, de-identification, rights, safe samples and controlled trials without receiving the original payload.
Usage Metering, Provider Settlement, Revocation Propagation and Audit Events
Measures RAG, API, evaluation and cleanroom usage, connects evidence contribution to settlement, and propagates rights changes through audit-ready events.
Non-export AI Model Evaluation Data Room
Keeps source data and ground truth inside a scorer-only environment, runs buyer models under restriction, and exports only approved metrics and redacted error summaries.
RAG-only, Eval-only and No-train Runtime Rights Broker
Compares declared intent with real API, SDK, job and data-sink signals to enforce RAG, evaluation and training restrictions at runtime.
Revocation Propagation Across RAG Indexes, APIs and Caches
Propagates rights withdrawal, contract expiry and de-identification changes to vector indexes, APIs, caches, embeddings, samples and access credentials, blocking residual use.