Private and secure,
by construction.
Every Northpoint feature has to pass one test: can it be described as "private and secure" to a compliance officer at an enterprise buyer? The product is built around that line.
The product hands an auditor a signed packet, not a screenshot.
Your model. Your endpoint. Your bill. We never proxy your bytes by default.
The audit chain records deterministic actions. LLM calls live alongside, as advisory traces.
Eight controls.
Each is non-negotiable. Each shows up in the product before it shows up in marketing.
BYO key by default
Customers route through their own Anthropic / OpenAI / Bedrock / Vertex / Azure OpenAI endpoint. Northpoint never proxies your bytes through our infrastructure unless you explicitly opt in.
PII redaction before LLM send
Schemas, distributions, and redacted samples ok; raw rows require explicit per-call consent. We index schemas — not raw cell values.
Audit every call
Prompt fingerprint (sha256), model id, token cost, response fingerprint, decision id. Logged alongside the deterministic actions they advise.
Tenant isolation, default-deny
RAG indexes are organization-scoped. Tenant scoping is the first WHERE clause. Per-tenant worker isolation available for BYO-key tenants.
On-device embedding option
High-security customers who can't send anything to a cloud model run all-MiniLM-L6-v2 in a sidecar process. The semantic index stays local.
Signed evidence packets
Every break investigation, every recon run, every agent proposal carries the lineage hash chain at the moment it was produced. Production deployments sign with the customer's ED25519 key.
Operator-guarded, always
Agents propose, humans accept. A proposal never mutates state without an explicit operator click. LLM calls are advisory traces, not actions.
Purge on request
Tenant data is purgable. Lineage rows, embeddings, audit traces — removable on a documented procedure. The deletion record itself enters the audit trail.
One LLM call,
seven verifiable steps.
Every copilot follows the same arc. Whatever leaves the box is in the audit trail, with a sha256 fingerprint of the prompt and the response.
- 1Operator actionOperator triggers a copilot — e.g. 'explain this finding'.
- 2Context assemblyNorthpoint reads lineage, profile, similar incidents — all locally.
- 3PII redactionCell values pass through the redaction filter. Schemas + distributions + redacted samples remain.
- 4BYO-key sendRequest is routed through the customer's LLM endpoint, with a per-call audit row written.
- 5ProposalLLM returns a structured proposal. Northpoint validates against the rule catalogue.
- 6Operator reviewOperator sees the proposal with citations; one click accepts as a deterministic action.
- 7Audit rowThe deterministic action enters the hash chain. The LLM call is logged as an advisory trace alongside.
Three deployment modes.
Customers pick the model that fits their compliance envelope. The product is identical across modes; only the boundaries move.
Hosted by us. Your data in your tenant; per-org keys; tenant-scoped RAG; audit log shipped to your S3 / GCS / Azure Blob.
Deployed into your VPC. Your network boundary; we manage upgrades over a control-plane that has zero data-plane access.
On-prem deployment. On-device embedding model; no outbound LLM by default; BYO LLM endpoint inside your perimeter.
Send your security questionnaire.
We’ll respond with the answers, the artifacts, and the diagrams your compliance team needs.