We change this documentation frequently to show the current state of the architecture.
Key point: Prem API is available today with full end-to-end encryption, CPU + GPU attestation, and a complete AI capability set (chat, audio, vision, documents, search). The main gap at launch is image reproducibility. Image reproducibility lets you rebuild the Prem enclave image from source and verify that it is identical to the production image. This gap is the top priority on the roadmap. All other items in the lists below are available now.
What Is Available Today
Encryption & Privacy
All features below are ready for production and available now.AI Capabilities
All standard AI features, with full encryption:Attestation
Hardware-backed proof that the correct code runs on genuine hardware:Platform & SDK
What Is Not Available at Launch
We tell you about the gaps clearly. Work on these features continues, but the features are not ready at launch.Image Reproducibility
What it is: Image reproducibility lets each person rebuild the exact enclave image from the published Prem source code. The person can then verify that the binary is identical, byte for byte, to the binary in production. Why it matters: Today, attestation proves that the enclave runs a specific image (the fingerprint agrees with a published value). But you cannot verify that the published fingerprint agrees with the open-source code. You trust the Prem build process. Reproducible builds close this gap. You build the image yourself, compare the fingerprint, and verify the match independently. The current status:- The build infrastructure is ready (Packer-based, pinned base images, pinned driver versions)
- The build pipeline produces consistent images
- You can verify attestation fingerprints against the published Prem values
- You cannot yet reproduce the exact binary independently, because we did not remove all sources of non-determinism (timestamps, file order, package manager variance)
- Deterministic build tools that remove non-determinism
- Published step-by-step instructions for independent reproduction
- Automated tools for hash comparison
- CI/CD integration that publishes reproducibility proofs with every release
Attestation Hardening
Incremental improvements to the attestation system. We release them continuously:Roadmap
The list is in priority sequence. We release features when they are ready, not on set dates.Next Priority
Reproducible Enclave Images
Deterministic builds with published instructions for independent verification. This completes the trust chain from source code → build → operational enclave. This is the most important gap for Prem to close.
Attestation Hardening
CRL checks, embedded root CAs, stricter TCB matching, and automated measurement verification. Incremental improvements that we release continuously.
Third-Party Security Audit
An independent audit of the platform is planned. We will publish the results here when the audit completes.
Medium-Term
Additional Self-Hosted Models
More model families on the Prem confidential infrastructure.
Enhanced Document Search
Improved embedding models, better document chunking, and cross-document search.
Longer-Term
Native SDKs for More Languages
Python, Go, and Rust SDKs with built-in encryption, in addition to the current local proxy method.
Verification Center
A visual interface to examine attestation reports, see measurement history, and monitor enclave health. It makes verification accessible to non-specialists.
Enclave-to-Enclave Chaining
Multi-hop confidential computing in which enclaves attest to each other. This permits complex confidential workflows across services.
The Trust Progression
The Prem API trust model becomes stronger in stages. Each milestone decreases what you must accept without proof: Today: Attestation proves that the enclave runs a specific, fingerprinted image on genuine hardware. You can verify the hardware signatures and confirm freshness. The gap: you trust that the published Prem fingerprints agree with the source code. Next: Reproducible builds let you build the image yourself and compare the fingerprints. This closes the trust gap between the source code and the operational enclave. Then you trust only the hardware manufacturers and the laws of mathematics. Then: A verification center makes continuous monitoring accessible to everyone. It supplies visual dashboards, alerts on fingerprint changes, transparency summaries, and historical audit trails. Each person in your organization can then check security verification, not only your security team.This page shows the current state of the platform. We update the page when the platform changes. For questions about specific features or schedules, send an email to support@premai.io.