NEAR AI Unveils Confidential Computing Platform for Secure AI Model Execution
NEAR AI is building infrastructure to safeguard sensitive data sent to artificial intelligence models for inference. The confidential computing platform, tied to the NEAR Protocol ecosystem, uses Trusted Execution Environments (TEEs) to protect both open-weight and closed AI models from unauthorized access.
The platform integrates Intel TDX and NVIDIA Confidential Computing technologies, which create encrypted memory spaces that shield code and data from the host operating system, cloud provider, and NEAR AI itself. Cryptographic attestation verifies that the correct model ran in a genuine secure environment without tampering.
NEAR AI Cloud offers an API for private inference, supporting both open-weight models like Llama or Mistral, which can be loaded directly into a TEE, and closed models from providers who don't release weights, accessible through a proxied setup. The platform has partnered with Venice.ai, Brave Nightly, and the Bermuda government to enable fully private inference for various applications.