arrow_back All work
AI Governance for Government ministries & regulated enterprises

Koine

The safe pipe between the data an institution already holds and an AI model — so regulated bodies can finally mine their own databases with AI, without ever risking a leak or an unauthorised change.

What it isGovernance-first AI-to-database middleware
Timeline2025 — present
StatusLive
Propose → approve
no unauthorised database change, ever
100% auditable
every action logged, attributed & reversible
PII stripped
automatically, across every interface
63 tests
governance verified end to end
How it works
Koine — how it works
target The challenge

Regulated organisations want to point AI at the databases they already run on, for instant insight — but wiring a model straight to live data is dangerous: unmonitored changes, personal-data leaks, and edits no one can trace or undo.

build What we built on their data

We built Koine, a governance layer that sits between the existing database, the AI and the people using it. The AI can only read and propose — every change a human must approve before it happens, and every personal detail is removed automatically across chats, reports and exports. Nothing happens that isn't logged, attributed and reversible.

Koine is the governance layer that makes AI-on-your-own-data safe: the AI can read and propose, a human approves every change, personal data is stripped automatically, and nothing happens that isn't logged, attributed and reversible.

Inside the system
AI governanceRead-only by defaultHuman-approval workflowAutomatic PII redactionFull audit trail
Built with
DjangoPostgreSQLLLM-agnostic
Visit Koine open_in_new
$ ./start

Sitting on data like this?

Tell us what you already run on. We'll show you the value hiding inside it — and scope a realistic engagement.

Book a consultation arrow_forward