Nillion is built on a simple belief: data can be used without ever being revealed. The Blind Computer makes that real, a decentralized system where data is stored, shared and computed on without being exposed.
Since the mainnet launch, the vision is turning into reality. With the Phase 0 upgrade, we shipped production-ready privacy, decentralized storage and AI. In just months, the network has already stored nearly 500 million secrets and processed more than 1.5 million AI inferences, powering applications across different verticals.
Now, with Phase 1, the Blind Computer takes its next step. This release introduces nilCC, unlocking general-purpose confidential compute, alongside major upgrades to nilAI and nilDB. Together, these advances make the Blind Computer’s foundation stronger than ever, enabling developers to build applications that unite private storage, private inference, and confidential compute, all backed by verifiable attestations.
nilCC: Confidential Compute for Everyone
nilCC is Nillion’s confidential compute layer, designed to let developers run sensitive workloads inside Trusted Execution Environments (TEEs) while producing attestations that verify integrity and authenticity.
In Phase 1, nilCC becomes fully usable for builders, with features that make deploying confidential workloads simple, transparent, and verifiable:
- Workloads via Docker Compose: Define workloads in Docker Compose and deploy them directly through the nilCC API. To simplify usage, a workload manager tool is available for creating, uploading, and managing deployments. Check out the docs here.
- Attestations: Every workload produces an attestation report with two critical components: verification of the TEE’s authenticity, and a measurement hash that confirms the workload matches the intended open-source code repository. This ensures what’s running is exactly what was expected.
- Open Source: All core components of nilCC are open sourced, ensuring transparency in how workloads are executed. You can refer to the repo here.
Developer Access: nilCC is live on mainnet and ready for production use. However, during this ramp-up phase, developers will need to contact the Nillion team to access credits for running workloads. Apply for access here.
This release marks the first step toward general-purpose Blind Compute. From sensitive business logic to AI workloads requiring strict confidentiality, nilCC provides a secure, verifiable foundation.
And it’s already being put to use with nilAI, the first product built on nilCC, unlocking the true power of confidential compute.
nilAI: Better Models, Stronger Privacy
nilAI, Nillion’s private inference layer has expanded bringing new models, built-in web search, and direct integration with nilDB to keep system prompts private.
What’s new:
- System Prompt Storage: For the first time, developers can own their system prompts as private assets. They can be stored securely in nilDB, shared through NUCs, and used directly in nilAI without exposing the underlying content. This enables prompt monetization and protects intellectual property, a novel capability we’re especially excited to share and a use case we plan to explore further in the future.
- New Models: nilAI supports Gemma 27B (multimodal with image input) and GPT OSS 20B, alongside the existing Llama8B model . Builders now have broader options across performance, and cost.
- Privacy-First Web Search: Built-in web search is now available in nilAI through Brave, chosen for its privacy-first approach that aligns with Nillion’s vision. The feature is off by default, and when turned on, users are clearly notified that queries may be visible to Brave servers. Watch web search in action in nilGPT.
With the latest developments, nilAI has evolved into a broader privacy-native AI environment.
As compute and AI evolve, access must also become stronger, Phase 1 strengthens it through nilDB.
nilDB: Decentralized Access in the Browser
nilDB will become a lot more powerful by the end of 2025, bringing decentralized access control directly into the browser.
What’s new in nilDB:
- Browser-Signed NUCs: Permissions can soon be granted, revoked and verified directly in the browser through MetaMask. No server-side signing or complex key management will be required, this reduces the complexities of building with NUCs and makes integration for consumer apps seamless.
- Pagination Implemented: nilDB will soon support pagination in its API, simplifying how builders retrieve secrets and helping them build faster and reliable products.
The Road Ahead
Phase 1 transforms the scope of the Blind Computer. The network has become significantly stronger, with confidential compute unlocked and core modules upgraded to support workloads far beyond storage or inference.
The Blind Computer opens the path to that future, where developers build privacy-first applications seamlessly, and users own their data by default.
Ready to build on Nillion? Follow the Quickstart guide and get started today.