nFOX Plans to Open-Source Blockhouse AI Sandbox for Enterprises

nFOX Lab plans to release Blockhouse as open-source software, pairing an AI workload sandbox with a newly launched platform for managing open models on infrastructure controlled by enterprise customers.

The nFOX Platform is designed to evaluate, deploy and monitor open models. Blockhouse is intended to limit what a model workload can do on the systems where it runs. The company presents the products as tools for organizations operating open models on infrastructure they control.

Blockhouse applies workload restrictions at the Linux kernel level. Model files are sealed with SHA-256 hashes, and serving halts when a file change is detected. File-integrity checks are part of the approval process before deployment.

Model traffic passes through Porthole, an audited communication channel. Runtime credentials and audit controls stay outside the model process. The design keeps those controls separate from the process running the model.

Blockhouse is designed for customer-selected cloud virtual machines, rented GPU hosts and on-premises Linux servers equipped with Nvidia GPUs. It supports vLLM, SGLang and llama.cpp. The platform’s approval workflow includes configuration reviews, file-integrity checks and adversarial-prompt testing.

For production workloads, the platform supports OpenAI-compatible API endpoints. That lets teams use a familiar interface to connect with deployed models, while Blockhouse provides the workload boundary. Together, the tools cover model management and runtime restrictions for organizations running open models on their own infrastructure.

Blockhouse does not replace trusted infrastructure or sound operating practices. The company announced the platform launch on Oct. 15. Blockhouse is in beta for qualified enterprise evaluations, and users can request access to its source code.