Skip to main content

Key Concepts

This glossary defines key terms as they are used throughout the nilAI documentation and system.

Access Modes​

There are 3 access modes available for using nilAI:

  • API Key mode is the plain, straightforward, string key that can be used as a Bearer Token
  • Direct Signing used with the nilAI SDKs takes advantage of Nillion's NUCs and consists of a Public DID & Private Key
  • Delegated Signing is similar to the above, but used to provide permissions to another users

Attestation​

A cryptographic proof mechanism that verifies the integrity of the TEE environment. nilAI provides two layers of attestation:

  • Environment Attestation: Verifies the GPU environment and service integrity through a verifying key
  • Response Signatures: Cryptographic signatures included with every chat completion response, proving the response originated from the attested environment

NVIDIA Confidential Computing​

The hardware security technology that enables TEE-based AI inference in nilAI. NVIDIA Confidential Computing provides GPU-level isolation and encryption, ensuring that AI model processing remains private even from the infrastructure provider.

OpenAI-Compatible API​

The interface standard used by nilAI for AI inference. Existing applications using OpenAI-style code can migrate to private LLMs by changing the endpoint URL while keeping the same code structure and API calls.

Signed Responses​

Cryptographic signatures included with every chat completion response from nilAI. These signatures allow verification that:

  • The response came from an attested TEE environment
  • The response has not been tampered with
  • The processing occurred in a verified secure enclave