Corvex Launches Secure Model Weights, Delivering Hardware-Enforced Protection for AI Inference on Third-Party Infrastructure

Patent-pending solution gives AI model builders and security-conscious enterprises cryptographic, owner-controlled key custody, protecting model weights from even the infrastructure provider

Mar. 13, 2026 at 2:53am

Got story updates? Submit your updates here. ›

Corvex, an AI infrastructure platform, has announced the early availability of Corvex Secure Model Weights, a patent-pending solution that enables AI model builders and enterprises to deploy inference workloads on third-party GPU infrastructure without exposing their model weights, which can be their most valuable intellectual property. The solution leverages hardware-based Trusted Execution Environments (TEEs) to ensure that model weights remain cryptographically isolated and are only decrypted within the GPU's secure silicon boundary, rendering them invisible even to the infrastructure provider.

Why it matters

Frontier AI models often represent years of research and hundreds of millions of dollars in compute investment. Any organization fine-tuning models on proprietary data is embedding sensitive IP directly into model weights. Traditional cloud security models leave data at runtime exposed, creating a critical vulnerability. Corvex Secure Model Weights addresses this risk by ensuring that model weights remain cryptographically isolated and are only decrypted within the GPU's secure silicon boundary, protecting them from even the infrastructure provider.

The details

Corvex Secure Model Weights combines three integrated technologies to enforce protection at the silicon level: Trusted Execution Environments, Remote Attestation, and Post-Quantum Key Exchange. The solution uses the open source Confidential Containers (CoCo) project under the Cloud Native Computing Foundation as its orchestration layer, providing vendor-neutral, community-audited security that customers can independently verify.

  • Corvex Secure Model Weights enters early availability on March 12, 2026.

The players

Corvex, Inc.

An engineering-led AI infrastructure platform specializing in GPU-accelerated infrastructure for AI workloads and purpose-built for secure, high-performance compute.

Seth Demsey

Co-founder and co-CEO of Corvex.

Jay Crystal

Co-CEO and Co-Founder of Corvex.

Got photos? Submit your photos here. ›

What they’re saying

“Deploying AI should never require a trade-off between compute power and IP sovereignty. The industry has long tolerated a 'cleartext gap' where weights are exposed during inference, leaving the host with a structural window into your trade secrets. We are closing that window.”

— Seth Demsey, Co-founder and co-CEO of Corvex

“Model builders and security-conscious enterprises are now able to choose infrastructure partners based on verifiable security, not just price and availability. An open source foundation and owner-controlled key custody are what make that security auditable and trustworthy.”

— Jay Crystal, Co-CEO and Co-Founder of Corvex

What’s next

To learn more or request early access, visit www.corvex.ai/confidential-computing.

The takeaway

Corvex's Secure Model Weights solution provides a critical layer of hardware-enforced protection for AI model weights, enabling AI model builders and security-conscious enterprises to deploy inference workloads on third-party infrastructure without exposing their most valuable intellectual property.