"What data did that model touch?"
Nobody can answer with evidence, only with intent.
NOOSChain is a permissioned blockchain for governed data sharing. Per-record permissions and encryption. Policy-driven residency. Compute-to-data workloads that bring approved scripts and AI models to the data's home node — and record verifiable proof of everything that happened. An industrial-grade trust layer for AI and data ecosystems.
You run models on data you don't fully own. You share datasets with partners you can't audit. When the regulator asks a question, you answer with screenshots. The failure modes are familiar:
Nobody can answer with evidence, only with intent.
Data sharing projects deadlock on who hosts the central platform and who sees what. The "who do we trust to run this" question kills multi-party initiatives before they start.
Either the partner gets the whole dataset, or the collaboration doesn't happen.
Copies proliferate to satisfy every team. Residency and retention obligations become unanswerable.
EU AI Act-adjacent requirements are coming, and your current tooling proves nothing.
Governance is a property of the infrastructure, not a wiki page.
One verifiable history of data, ownership and activity across organisations. No need to appoint one of you as the trusted central party.
Membership, permissions and state changes are explicit, signed and auditable from the start. Governance is a property of the infrastructure, not a wiki page.
Chain state is separated from encrypted payloads. Per-record and per-bucket encryption, four-tier sharing (public / owner-only / org-shared / group-shared). Share exactly as much as needed, no more — avoiding both over-sharing (privacy/IP risk) and under-sharing (collaboration friction).
Replication policies determine where data may travel, where it must remain and which nodes may retain it. GDPR and data-sovereignty constraints become configuration.
The decisive inversion: instead of exporting data to the model, you send the signed model to the data. Execution results, artifacts and provenance are recorded on-chain. The data never leaves its authorised environment. Consensus orders work; execution applies it. Crash-fault-tolerant by design.
Package an analytical or AI workload as a signed script or container image. NOOSChain authorises it against specific buckets, routes it to eligible data-holding nodes, and executes it inside their local environments.
Publish a signed workload with an immutable script or container-image digest.
Approve its identity, buckets, purpose, runtime limits and output policy.
Select nodes that hold the required data and are permitted to execute it.
Query protected buckets inside the node's controlled execution environment.
Commit execution status, results, artifacts and provenance to the blockchain.
Authorising a workload does not authorise its creator to read the underlying data. Access is scoped to the signed workload, selected buckets, eligible nodes and its execution policy.
Read the guideThe signed workload — script or container reference, digest and execution policy.
Protected bucket data — raw records remain inside their authorised environment.
Verified execution — status, result, artifacts, hashes and provenance.
From model governance to data marketplaces — every outcome below shares the same verifiable execution trail.
Independently verifiable proof of what data a model was trained or run on, when, and under whose authorisation. Audit-ready answers for regulators, partners and internal risk — without disclosing the data itself.
Train or fine-tune inside authorised data environments with controlled model-artifact retention, when the data cannot leave institutional or organisational walls.
Run permitted statistics, features and derived-dataset extractions across partners' data while every source record stays protected. Share the insight, not the data.
License or share access to specific datasets under defined terms. The granular permission model means the data owner retains ownership. The buyer gets exactly the access they paid for — and the chain proves they got nothing else.
Compliance, quality and transformation workflows with a verifiable execution trail, addressing emerging AI-accountability requirements without full data disclosure.
Each layer has one job: applications express intent, validators agree on order, deterministic execution updates state, and authorised compute nodes run data-local workloads with verifiable results.
Build and sign transactions — signed intent, via TypeScript SDK or HTTP API.
Reach consensus, order and commit transactions — deterministic execution.
Verify state, control data — state roots, snapshots, replay for node recovery.
Every layer has one job. The whole system is a deterministic state machine: given the same ordered transactions, every validator reaches the same result — and can verify it.
Understand the architectureNOOSChain is our product. We designed it, we run it in production across our own EU projects, and AI governance is exactly why we built it. The Web UI is live right now: real chain height, live blocks, a data explorer. See the chain move before you talk to anyone. And when you're ready to put real data under governance, you don't have to figure it out alone. Bring us your pipeline and we'll design the network with you — permissions, residency, first workload. Direct access to the people who wrote it.
Book a meetingStart NOOSChain with Docker and inspect the node using the CLI and Web UI.
npm run docker:up
The Web UI is public right now — real chain height, live blocks, a data explorer.
Open the Web UIPermissions, residency, first workload — designed with the people who wrote it.
Book a meetingNoosware's roots are in machine learning, robotics and cloud systems; AI is the core of everything it builds — including LLM-application architecture proven in production products. NOOSChain is the trust layer underneath those AI pipelines: proprietary infrastructure used across Noosware's EU Horizon projects (BETTER, MARS, ICAERUS, RAISE Suite). Built by product people, not just researchers — scoped through discovery, validated early, designed for product-market fit. It is not a general-purpose blockchain, and there is no public token: this is infrastructure, not a speculation play.