PRISM Security Architecture

Your documents are mathematically locked to your specific user identity using strict Row Level Security (RLS) database protocols. It is architecturally impossible for another user, firm, or tenant to access, query, or even confirm the existence of your files. This is not a permission setting — it is a mathematical constraint enforced at the database layer.

Your documents and chat history are never used to train any AI model. PRISM is powered by Microsoft Azure Enterprise AI infrastructure — the same security and compliance standards trusted by the world's largest legal and financial institutions. Azure operates with Content Logging disabled. The system processes your document in memory to provide citations, and the models retain nothing. Microsoft cannot see your data. Neither can we.

In transit: all data is transmitted over strictly enforced HTTPS/TLS 1.3 connections. At rest: every file, text chunk, and vector embedding is encrypted using AES-256 — the same standard used by global financial institutions and governments. Encryption is not optional and cannot be disabled.

You retain absolute control over the lifecycle of your data. With a single click, you can execute a permanent deletion. This action immediately and irreversibly purges the original PDF, all generated text chunks, all vector embeddings, and the complete associated chat history. A timestamped Destruction Receipt is generated as cryptographic proof of permanent deletion — a legal artifact you can retain for compliance purposes.

Every answer PRISM generates is accompanied by a direct citation to the exact page and paragraph of your original document. You never have to trust the system blindly — you can verify every claim instantly against the source material. This is not a feature. It is the architectural foundation of the system.