NEAR AI unveils cloud and private chat solution that puts users in control of AI

by
Alyssa Davidson

Posted: December 4, 2025 3:38 AM Updated: December 4, 2025 3:38 AM

by well

Correction and fact check date: December 4, 2025, 3:38 AM

We sometimes use automatic translation plugins to improve your local language experience. The automatic translation may not be accurate, so please read the original text for accurate information.

Modular Protocol for AI Near has launched NEAR AI Cloud and Private Chat, describing it as a product built around the principle that users should retain control over AI systems. The announcement positions these services as an initial step toward enabling user ownership through mechanisms designed to provide hardware support and verifiable privacy.

The company noted that while AI capabilities are advancing rapidly, most AI workflows still do not allow secure transfer of sensitive and proprietary information. Many existing AI APIs require data to be sent to external infrastructure that cannot be audited by users, and the commonly advertised private modes typically do not provide isolation, restricted access, or verifiable evidence of data deletion. Using an internally hosted system can reduce exposure, but often requires significant resources, expertise, and a long deployment period.

According to the announcement, existing trust mechanisms such as software controls, internal policies, contractual agreements, and compliance certifications provide only indirect assurances and can be compromised due to errors or internal misuse. These limitations are cited as a reason why many organizations avoid exposing valuable data to AI systems, and as evidence of a broader industry need for privacy solutions that can be verified rather than assumed.

The NEAR AI suite is presented as an attempt to bridge this gap. NEAR AI Cloud processes each request within an Intel TDX and NVIDIA confidential computing environment. Here, data is processed in an isolated and sealed context that cannot be accessed by service providers or hardware operators.

Each inference generates a cryptographic proof that the model ran on verified hardware using the expected code, and these proofs can be independently verified using public verification services provided by Intel and NVIDIA.

NEAR Private Chat powers on NEAR AI Cloud and applies these privacy guarantees to everyday conversational use cases, including informational, personal, or plan-related queries. It is described as a service intended to provide verifiable privacy so that everyday interactions with AI can occur without exposing user conversations or related data.

Now deployed in production using Brave Nightly, OpenMind, and Phala

NEAR AI Cloud is currently used by partners including the Brave Nightly browser, robotics operating system provider OpenMind, and confidential cloud platform Phala. These deployments cover a wide range of privacy-sensitive situations, from consumer interactions to autonomous system operations, all requiring verifiable privacy. These platforms reach more than 100 million users globally across consumer and enterprise domains.

NEAR AI has worked directly with the engineering and security teams of these organizations to test confidential computing against real-world production requirements, including high traffic volumes, predictable latency requirements, and stringent compliance expectations. Input from these partners has influenced platform development, resulting in a system that is already proven reliable and capable of supporting sensitive workloads at scale.

NEAR AI Cloud is designed to simplify the process of transforming existing AI workloads into a private, verifiable environment. Its API remains fully compatible with the OpenAI interface, so your current integrations can work without modification. Python and TypeScript SDKs provide native support for encryption, key handling, and attestation processes.

Operational performance remains stable, with expected latency increases of approximately 5-10% and support for request rates of up to 100 requests per second per tenant. Dedicated and custom model hosting options are available for enterprise scenarios, with features such as streaming, key-value caching, and model persistence.

The platform aims to make confidential computing accessible with standard cloud capabilities while emphasizing ease of adoption, scalability, and production readiness. During the first half of 2026, NEAR AI plans to introduce multimodal models and expand platform capabilities beyond inference for additional data-driven applications, including initial work on private portable memory and continued progress on the DCML roadmap.

Our ongoing product direction is focused on enabling verifiable privacy as a fundamental characteristic of all AI workloads, whether everyday interactions or large-scale enterprise systems.

As MPost’s resident journalist, Alisa specializes in the broad areas of cryptocurrencies, zero-knowledge proofs, investing, and Web3. With a keen eye for new trends and technologies, she provides comprehensive coverage to inform and engage readers about the ever-evolving digital financial landscape.

• 
  
  

more articles