Ultraviolet Logo

confidential-computing Prism AI Public Beta: Empowering Secure AI Collaboration

sammy oina

sammy oina

· 4 min read

Prism AI Public Beta: Empowering Secure AI Collaboration

In an era where AI is transforming industries, the challenge of data privacy remains a significant hurdle. Organizations often find themselves choosing between the power of state-of-the-art AI and the security of their most sensitive data.

Today, we are thrilled to bridge that gap with the launch of the Prism AI Public Beta.

Prism AI is a web-based SaaS platform designed to make Confidential Computing accessible, powerful, and truly collaborative. It provides an enterprise-ready interface for orchestrating zero-trust collaboration via Secure Multiparty Computation (SMPC), powered by Trusted Execution Environments (TEEs).

Through the technical guarantees of hardware-level isolation, Prism AI allows multiple organizations to jointly train models or run private inferences in a shared TEE. Critically, raw datasets and proprietary algorithms are technically shielded from every other party in the computation—including collaborators and the cloud provider itself—ensuring absolute privacy throughout the entire lifecycle. Built on the open-source Cocos AI foundation, Prism AI turns trust from a policy into a physical law.

Why Prism AI is a Game-Changer

Traditional AI workflows require you to trust your infrastructure provider. Whether you’re using a public cloud or a third-party API, your data is visible to the underlying system. Prism AI changes this paradigm by moving from policy-based security (“we promise not to look”) to technology-based security (“it is physically impossible for us to look”).

The Anatomy of a Technical Guarantee

Prism AI isn’t just a layer of software; it’s a bridge to hardware-level security. Here is how the platform ensures your data remains confidential:

  1. Hardware-Level Isolation (TEEs): Prism AI orchestrates Confidential VMs (CVMs) powered by AMD SEV-SNP or Intel TDX. This means your data and AI models are encrypted in the system’s RAM. The decryption keys are managed by a secure, hardware-embedded processor that is inaccessible even to the cloud provider’s core operating system or hypervisor.

  2. The TEE Manager: Running on host hardware, the TEE Manager is a critical open-source microservice that dynamically provisions and configures the secure enclaves. Once a computation is finished, it ensures the TEE is securely destroyed, leaving no trace of the sensitive data behind.

  3. Attested TLS (aTLS) & The In-Enclave Agent: Before any data or algorithm is uploaded, Prism AI performs a Remote Attestation handshake. The In-Enclave Agent (running inside the secure VM) provides a cryptographic “quote” signed by the hardware. Prism AI verifies this quote against strict security policies. Only after the hardware’s identity and software’s integrity are proven is a secure aTLS tunnel established for data transmission.

  4. Policy-Driven Orchestration: Prism AI’s control plane ensures that computations only run if every participant’s security requirements are met. This moves the trust model from human promises to high-assurance, mathematical proofs.

Built on an Open Source Core

Transparency is fundamental to trust. That’s why the core components of Prism AI—including the TEE Manager and the In-Enclave Agent—are part of the open-source Cocos AI project. This ensuring that any organization can benefit from robust TEE orchestration, whether in the cloud or on-premises.

Real-World Applications

Prism AI is already being used to solve critical privacy challenges in high-stakes industries:

  • Healthcare: Training diagnostic models on patient records across multiple hospitals without violating GDPR or HIPAA.
  • Finance: Collaborative fraud detection and risk assessment across institutions without sharing proprietary data.
  • Governments: Enabling secure data sharing between agencies for public safety and research while maintaining strict data sovereignty.

Join the Mission for Secure AI

We’re on a mission to make privacy the default for artificial intelligence, and we’d love for you to be a part of it. The Prism AI Public Beta is officially live, and we can’t wait to see what you build.

  • 🚀 Get Started for Free: Sign up at cloud.prism.ultraviolet.rs and start running your first confidential computations.
  • 📚 Explore the Docs: Dive into the technical details at docs.prism.ultraviolet.rs.
  • 🤝 Join the Community: Follow our progress on GitHub and help us shape the next generation of privacy-preserving AI.

The era of choosing between powerful AI and strict privacy is finally over. With Prism AI, you can have both—fearlessly.

Learn More: