NEAR AI Cloud launches private uncensored AI for cybersecurity

NEAR AI Cloud has launched a cybersecurity offering built around what the project describes as private, uncensored AI models running inside secure enclaves. The company said the service is live and aimed at security teams that want to test vulnerabilities, review code and work with sensitive findings without exposing that material outside the environment.

In a product page for the cybersecurity launch, NEAR AI says the models are designed to keep prompts, code and results private while removing refusal behavior from the model itself. The company also said the service is accessible through NEAR AI Cloud and can be used with an OpenAI-compatible interface.

NEAR AI said the launch includes models such as DeepSeek 4.1 Flash and Qwen3.8 27B, both described as “abliterated” in the company’s wording. The project added that the service is open to every account with no application or identity checks, and that pricing starts at $0.04 per million input tokens.

A security-focused rollout, not a general AI release

The announcement frames the product as part of a broader security workflow rather than a general-purpose chatbot release. According to NEAR AI, the models are meant to help defenders ask questions such as how code could be hacked, while keeping the underlying vulnerabilities private inside secure enclaves.

Illia, a NEAR co-founder, said the rollout is intended for security scanning and that all code, prompts and findings remain private within the enclave environment. NEAR AI also said it is pairing the launch with NEAR AI RED, a program for security teams that includes access to new models and workshops.

The company’s public materials say the service is live now. Beyond that, the launch details focus on access mechanics, privacy claims and the security-use case, rather than on broader adoption metrics or market impact.