The internet today doesn’t protect us, it extracts from us.
Our data is constantly observed, monetized, and fed into systems we can’t control.
Crypto challenged that model by proving that trust can be replaced with math. It enabled open, ownerless systems. But it prioritized transparency, not privacy, and the question remained: how do we make data usable without making it visible?
The Blind Computer makes that possible.
It enables private data to be stored, processed, and analyzed, without ever being revealed.
This is the heart of the Nillion Standard: building systems where you can use your data without giving it up. Shaping an internet designed for intelligence.
And the most radical part? It works.
From Vision to Mission
Our mission is making privacy viable in production, not just for demos or small use cases, but for real-world applications at scale.
Nillion’s Blind Computer has three core components:
- nilDB – the decentralized storage layer enabling terabytes of data to be stored securely among the network nodes.
- nilCC – the confidential general compute layer powering state-of-the-art AI (nilAI), DNA analytics and more.
- nilAuth – the authorisation layer that provides a decentralised permissioning for our storage and compute.
These components are designed to work as one system, a new stack where privacy is verifiable and programmable.
Leading enterprises are exploring how it changes what’s possible in their industries. Ecosystem partners are storing secrets, running inference, and pushing our infrastructure into real-world use cases. Developers are building on it. Node operators are running nodes.
This roadmap outlines what is currently live and what’s coming next for the network. With every phase, we’re moving toward a fully private, programmable compute layer. This is the first functional Blind Computer, and we hope it sparks many more. Like any foundational shift, our goal is to set a standard others can build on.
Phase 0: Production-Ready Privacy (July 2025)
Blind Compute is already powering real-world applications. Phase 1 builds on that momentum, expanding enterprise deployment, unlocking decentralized permissions, and enabling private AI use with verifiability.
July Release Highlights:
- A 10-node nilDB network, including a 6-node enterprise cluster.
- nilCC supports state-of-the-art AI compute serving Llama 3B, 8B, 70B, DeepSeek 14B (with Qwen 14B & Gemma 27B support coming soon), and providing attestation reports that can be checked for integrity.
- Efficient secure RAG, allowing for a performant way to provide context to the model.
- Moving from a static JWT permissioning model to a UCAN-compliant architecture, enabling decentralised, extensible, fine-grained access control.
Since January, several ecosystem projects have been actively using the Blind Computer, storing secrets, running production models, and providing continuous feedback that will continue to shape this phase.
To date, over 750M secrets have been stored in nilDB and over 1.2M inference calls made to nilAI.
This is what production-ready privacy looks like.
Phase 1: General Compute (Late 2025)
Nillion’s Blind Computer expands into general-purpose blind compute.
Whilst phase 1 proved that privacy and performance can coexist for storage and AI inference, phase 2 unlocks the ability for developers to run arbitrary workloads securely inside TEEs. Whether it’s sequencing DNA, running sensitive business logic, or executing an entire frontend in a TEE.
At the core of this phase is the deep integration between nilDB and nilCC. Developers will be able to run computation in nilCC that reads from and writes to nilDB, using private storage and compute in a single, programmable system. Privacy is enforced by the permissioning layer (nilAuth) and attestations output by nilCC. This functionality forms the backbone of Nillion’s Blind Computer, developers can utilize secure storage & general compute in one system.
We’re also enhancing our AI capabilities to support high performance use without compromising privacy.
Developers will be able to run more powerful models like Gemma 27B and Qwen 14B with full privacy and no data leakage. We will introduce private web search, supercharging model capabilities by letting them search the internet, without revealing identity, intent, or queries. And also implementing safety tools like the Llama Firewall, enforcing output filtering and run red-team tests without ever exposing inputs or results.
These improvements ensure nilAI evolves alongside the most advanced AI systems, while staying rooted in privacy by design.
Phase 2: Default Privacy UX (Early 2026)
With phase 0 and 1 complete, the Blind Computer is fully battle-tested and ready for large-scale, real-world deployment of private storage and computation.
By this stage, the infrastructure has proven itself in real-world conditions. Devs continue to build and scale, and privacy-preserving apps are no longer rare, they’re becoming the default.
Developer experience has always been a core focus, and phase 2 takes it even further. Building privately shouldn’t feel like a tradeoff. It should be just as seamless, fast, and intuitive as building without privacy at all.
This phase is about making Nillion’s Blind Computer the most exciting place to build, and where privacy-powered apps go mainstream.
The Privacy Renaissance has begun
The Blind Computer isn’t theory. It’s running. And with each phase, we’re proving that privacy and performance can co-exist together. Not as a feature, but as a foundation.
This is just the beginning. We’re building the missing layer of the crypto stack, one that brings privacy into the real world, at scale.
Let’s shape the future of privacy together.