NEAR AI Cloud Integrates Intel® Trust Authority: Independent Attestation for Confidential Inference

Confidential computing has a verification gap. Providers that run AI workloads inside trusted execution environments all make the same claim: code and data are protected in memory, and the machine operator cannot inspect them. The protection is a hardware property, but the proof usually is not. In most deployments, the party that operates the infrastructure is also the party that checks the attestation evidence. The operator attests, the operator verifies, and the customer is asked to accept the result.

For enterprises moving regulated data into AI systems, that circularity is the difference between a security property and a security promise. Compliance programs built on GDPR, HIPAA, or SOC 2 do not run on promises. They run on evidence that an independent party can check.

NEAR AI Cloud now closes that gap with Intel Trust Authority, Intel's independent attestation service. Confidential inference on NEAR AI Cloud is verified not only by NEAR, but by Intel, and every verified workload returns an Intel-signed attestation token as portable, independent proof.

What NEAR AI Cloud already proves with Intel® Trust Domain Extensions (Intel® TDX)  

Requests to TEE-hosted models on NEAR AI Cloud run inside an Intel TDX trust domain: a hardware-isolated confidential virtual machine whose memory is encrypted and whose contents are shielded from the hypervisor and the host. The host operating system, the GPU operator, and NEAR AI itself are cryptographically locked out. Each request returns a hardware-signed attestation.

NEAR AI Cloud validates the Intel TDX quotes underlying each attestation and publishes the raw evidence, so anyone can examine the measurements directly rather than rely on a summary. You don't need to take NEAR's word for the environment handling your data. The proof has always been there.

What stayed constant, as it does across every confidential computing provider, is that the primary verification path ran through the operator. That's the part that's now changing.

What Intel Trust Authority adds

Intel Trust Authority is an attestation verifier that operates independently of NEAR and of the cloud providers underneath. With this integration, evidence from a NEAR AI Cloud trust domain is sent to Intel Trust Authority, which checks the cryptographic signatures on the quote, appraises the trusted computing base against Intel's reference values and policy, and issues an attestation token: a signed JWT carrying its verdict.

The token does not assert privacy by declaration. It proves the workload is running inside a genuine, unmodified Intel TDX trust domain whose trusted computing base matches Intel's expected values. Confidentiality follows from that verified state: an environment proven intact is an environment its operator cannot read.

The verifier itself is inspectable. Intel Trust Authority's verification services run inside Intel SGX enclaves, and its Faithful Verification feature produces an audit report with cryptographic evidence that the services which issued a given token were themselves unmodified at the time of issuance. The chain of trust no longer terminates at the infrastructure operator. It terminates at the company that manufactures the hardware enforcing the isolation, and even that link can be audited.

What an independent verifier changes in practice

An Intel-signed token is machine-checkable proof that other systems can act on. Key management services can release decryption keys only against a valid token. Access control and security platforms can gate workloads on it. Compliance teams can hand auditors third-party attestation evidence instead of a provider's self-assessment. Anything built on top inherits it: agents and apps running inference on NEAR AI Cloud, IronClaw among them, get that verifiability for their inference without engineering it themselves.

It also changes the shape of the trust argument. Confidential compute offerings typically ask customers to rely on attestation checked inside the same trust boundary that runs the workload. NEAR AI Cloud separates the two: the operator runs the workload, an independent authority verifies it, and the raw evidence stays available for anyone who wants to check both.

"Verifiability is the key requirement for user-owned AI and is essential for data sovereignty and security. Adding Intel ITA independent attestation to all NEAR AI inference separates the verifier from the hardware operator and provides a machine-checkable proof to the user. With Intel as a confidential hardware partner and now a verification partner, NEAR AI is setting the standard for private and verifiable AI." –Illia Polosukhin, Co-Founder of NEAR and NEAR AI

Private AI, in practice

Most of what the industry calls private AI is a policy artifact: a data processing agreement, a retention pledge, a compliance page. None of it binds the machine. The shift underway in confidential computing replaces the pledge with a property. Privacy becomes something hardware enforces and an independent party confirms, much as TLS moved transport security from promise to protocol.

That shift is no longer theoretical. Hardware isolation ships in Intel TDX. Independent verification ships in Intel Trust Authority. The attestation token is machine-checkable, so downstream systems can enforce it instead of trusting it. Major platforms, enterprises, and governments are now using NEAR AI for confidential, private inference and agents, including Brave, Venice AI, Abound, the Government of Bermuda, OpenMind, Corbits, and more.

Verify it yourself

Confidential inference with independent Intel attestation is available today on NEAR AI. 

Learn more about NEAR AI Cloud or dive into the technical documentation to get started.