Unlock your most valuable data for AI with verifiable, hardware-enforced privacy
Today NEAR AI is launching NEAR AI Cloud and Private Chat, two products built around one simple yet powerful principle: users should own their AI. Today we’re taking the first steps to enable that with novel products that offer hardware-backed, verifiable privacy.
Privacy Concerns Are Blocking Large-Scale AI Adoption
AI is accelerating, but sensitive and proprietary data still can’t move safely with it. Most AI APIs require sending sensitive and proprietary information to infrastructure the user cannot inspect, and “private” modes offer no proof of isolation, access, or deletion. Self-hosting avoids the risk but demands heavy investment, specialized teams, and long deployment cycles.
Traditional trust signals (software controls, policies, contracts, certifications) are just signals, but in reality they’re very breakable. One misconfiguration or insider action can break them instantly. It’s why enterprises continue to keep their most valuable data out of AI, and why the AI industry at large needs verifiable privacy, not unsupported promises.
“Everyone should own their AI. Today NEAR AI is taking a major step towards the vision of user-owned AI with the launch of private, verifiable AI products. Privacy and verifiability are required to maximize available AI context for users and businesses, with confidentiality and compliance thanks to hardware- and cryptography-based guarantees.”
— Illia Polosukhin, Co-founder, NEAR Protocol
NEAR AI: Privacy You Can Verify
The NEAR AI product suite is built to close the AI privacy gap. With NEAR AI Cloud, every request runs inside Intel TDX and NVIDIA Confidential Computing hardware where data is processed in a sealed, isolated environment inaccessible even to NEAR AI or the hardware owner.
Each inference generates a cryptographic attestation proving the model ran in genuine, verified hardware with the expected code and can be independently validated through Intel’s and NVIDIA’s public attestation services.
NEAR Private Chat runs on NEAR AI Cloud and applies these cryptographic guarantees to everyday use cases, from relationship advice to financial planning. Everyone deserves access to the benefits of AI without the risk that their chats and data will leak. NEAR Private Chat finally introduces verifiable privacy to your everyday questions and research.
Now Integrated in Production with Brave Nightly, OpenMind, and Phala
NEAR AI Cloud is serving live customers with privacy browser Brave Nightly, robotics OS provider OpenMind, and confidential cloud provider Phala. Their environments span privacy-critical consumer interactions to real-time autonomous systems, use cases that require provable privacy guarantees. Cumulatively these platforms serve over 100 million consumer and enterprise users across the globe.
Working directly with their engineering and security teams, NEAR AI has validated confidential computing under real operational conditions: high-volume traffic, deterministic latency, and strict compliance requirements. These design partners helped shape the platform, ensuring NEAR AI at launch is already proven, reliable, and ready for sensitive workloads at scale.
How It Works
NEAR AI Cloud is a step towards NEAR’s roadmap to Decentralized Confidential Machine Learning (DCML). It delivers private inference in 3 steps:
- Encrypt: A user’s prompt is encrypted locally at the edge and transmitted to a Confidential Virtual Machine (CVM) inside a TEE.
- Isolate: Processing occurs inside a Trusted Execution Environment (TEE) that prevents external access, including from NEAR. Inside this secure enclave, the model decrypts the input, performs inference, and re-encrypts the result before returning it.
- Verify: The user decrypts the response locally, ensuring full end-to-end privacy. They also receive a proof that they can verify independently. To verify the computation, developers can call: /v1/attestation/report?model={model_name}
This endpoint returns signed proofs from Intel TDX and NVIDIA TEE that can be validated directly against their attestation services, confirming that the computation occurred in genuine secure hardware running verified code.
Built on NEAR AI Cloud, Private Chat offers the same privacy guarantees with leading open-source models through a familiar chat interface.
As NEAR AI expands, our long-term vision is a decentralized confidential compute layer coordinated on-chain. As AI workloads scale, they create sustained demand for the underlying settlement, verification, and incentive mechanisms, positioning the NEAR token as the native asset that powers and aligns the economics of this emerging AI cloud.
For Developers: A Lift-and-Shift Upgrade
NEAR AI Cloud makes it easy to move existing AI workloads to a verifiably private environment. The API is fully OpenAI-compatible, so current client libraries and code paths work immediately upon migration. Python and TypeScript SDKs handle encryption, key management, and attestation out of the box.
Performance remains smooth, adding approximately 5-10% percent latency while scaling to 100 requests per second per tenant. Dedicated and custom model hosting is available for enterprise use cases, supported by features like streaming, KV caching, and model persistence.
NEAR AI enables confidential computing to become a readily accessible cloud service: simple to adopt, scalable to operate, and production-ready.
What’s Next for NEAR AI
In H1 2026, we’ll introduce new models that support multiple modalities. The NEAR AI Cloud platform will expand beyond inference to broader use cases around user data, including exploring private portable memory and continuing to build the DCML roadmap.
Our product roadmap is driven by a clear and ambitious goal: make verifiable privacy the default for every AI workload, from individual conversations to enterprise-scale systems. If you’re interested in shaping the future of user-owned AI, we’d love to work with you.
Get Started
- Read the API reference: https://docs.near.ai/api
- Read the documentation: https://docs.near.ai/
- Github: https://github.com/nearai
- Try Private Chat: https://private.near.ai/welcome
