An AI governance assessment holds sensitive detail about how an organisation operates. This page sets out how Aegis protects it, in plain terms.
Every table that holds an organisation’s data is protected by Postgres row-level security. Each request declares who is acting before it runs, and the database itself decides which rows that person may see.
The application connects as a role that owns no tables and cannot bypass these policies. A query written incorrectly still cannot return another organisation’s records.
The standard deployment runs on managed cloud infrastructure. The sovereign deployment is a dedicated, single-tenant installation in a Ghana data centre, so interview content and evidence stay in Ghana.
On the sovereign tier, interviews and drafting can run on self-hosted models, so no content needs to leave the country for AI processing.
Significant actions, including reviews, approvals and signatures, are written to an audit log the application can add to but not edit or remove. Interview answers are append-only in the same way: a correction is a new answer, and the original is kept. The one exception is a deliberate, recorded erasure when data protection law requires deletion.
Data is encrypted in transit between your browser and the platform, and our deployment standard requires encryption at rest. Per-organisation encryption keys for uploaded evidence are on the roadmap.
AI providers sit behind adapters, so a deployment can use a different provider or a self-hosted model without changing how scoring works. Our policy is to use AI providers only on terms that prohibit training on your data.
The division is enforced in how the platform is built. The AI produces language; rules and people produce decisions.
Uploaded documents and transcripts are passed to the AI as data to read, never as instructions to follow.
Every issued document carries a QR code and a verification code. The public check confirms the document type, the issuing organisation, its version and date, and whether it is current or superseded.
It reveals no score, finding or content. Drafts are never verifiable, because they were never issued.
Aegis does not yet hold third-party security certifications. An independent penetration test is planned before the first public-sector deployment goes live. We are happy to answer security questionnaires and share our data processing agreement on request.
A 45-minute walkthrough with a RAMSYS consultant: the interview, a scored assessment, and the documents it produces.
