mechanism both

Classification containers

Rules live on the data, not on each tool. Default-closed. Unclassified is refused. Every source signs a container before it is read. Tools that connect later inherit the same rules. We never hold your data in our cloud, so we cannot lose what we never have.

This node Classification containers

Today every AI purchase is its own security review — configured per tool, wrong per tool, forgotten per tool.

In Mnemma the class travels with the knowledge. Containers are RED or GREEN. Access is per identity. A tool that connects inherits the walls. Govern once.

RED serves only to principals whose model path is cleared for RED. A GREEN principal asking a RED question is denied — and that deny is a first-week test we run in front of you.

Claims

What this page stands behind.

#claim-default-closed

Classification is container-based and default-closed. Unclassified is refused.

Applies to
Standard and Greenfield
Evidence
Nothing connects without a declared container.
Since
2026-08-20
#claim-red-green

Data lives in declared RED or GREEN containers. Access is per identity.

Applies to
Standard and Greenfield
Evidence
The wall is on the knowledge, not on the client's promise.
Since
2026-08-20
#claim-inherit

Every tool that connects inherits those rules. Govern once.

Applies to
Standard and Greenfield
Evidence
Next year's tool does not get a new security novel. It gets the same walls.
Since
2026-08-20
#claim-red-path

RED serves only to principals whose model path is cleared for RED.

Applies to
Standard and Greenfield
Evidence
Trust is the inference path. The fully air-gapped, zero-egress path is the on-prem model.
Since
2026-08-20

Machine-readable → index.json · /ai-info · llms.txt

browse The rest of the brain