The AI your auditorwill love.
private·ai is a compliance-first assistant for regulated teams. Every retrieval is logged, every answer is grounded, every span is traceable — without sending a byte to the cloud.
Built by Safe4AI
private·ai is designed and engineered by Safe4AI — the same team that deploys on-premise AI infrastructure for regulated enterprises across healthcare, finance, legal, and government.
What's in the box
Built for the people who get subpoenaed.
Six capabilities, working as one system — so your team can ship AI without making your CISO file a JIRA ticket.
Every query, on the record.
Persistent audit rows for every prompt, retrieval, latency tick and trace ID. Exportable as CSV. Retained 90 days, archived to immutable storage.
Three filters between you and harm.
Input guard blocks injections and blocked terms. Content filter scrubs PII from retrieved chunks. Output filter checks for hallucinated PII and bounded length.
Hybrid retrieval, fused and reranked.
Dense vector similarity combined with sparse BM25 ranking, fused with RRF and reranked. Better recall, better grounding, fewer hallucinated answers.
OpenTelemetry, all the way down.
Full distributed tracing via OTEL spans, shipped to Jaeger. See where latency lives, why a retrieval missed, and which model answered.
Answers cite their own evidence.
Every assertion is anchored to the source chunk and page. When the corpus is silent, the model says so — instead of inventing a plausible lie.
Your network. Your data. Your weights.
Runs entirely in your VPC. Local Ollama for inference, Qdrant for vectors, Postgres for audit. Zero data egress. No API keys to a third party.
How it works
A pipeline you can defend in a deposition.
Seven discrete stages, every one observable and overridable.
User
auth · sso · rbac
Input Guard
injection · pii · policy
Hybrid Retrieve
dense · bm25 · rrf
Content Filter
pii redact · k filter
LLM Generate
ollama · qwen · vllm
Output Guard
pii · length · cite
Audit + OTEL
postgres · jaeger
Why not just use a cloud LLM?
Because "trust us" is not a compliance control.
See how private·ai compares to generic cloud LLM providers across the capabilities your compliance team actually cares about.
| Capability | private·ai | Generic Cloud LLM |
|---|---|---|
| Data leaves your network | Never | Every request |
| Audit trail | Persistent, queryable | Log on best effort |
| PII redaction in retrieval | Built-in content filter | DIY in your wrapper |
| Grounded citations | Required by default | Optional, often omitted |
| Model choice | Local · hybrid · cloud | Whatever the vendor ships |
| Distributed tracing | OpenTelemetry, every span | Vendor dashboards only |
| Tenant isolation | Your VPC, your weights | Shared infra |
| Cost per query | Predictable (your iron) | $$$ at scale |
† Cloud LLM positioning generalized from the top three vendors as of Q2 2026. Your mileage may vary; your auditor's tolerance will not.
Ready when you are
Bring your hardest compliance question.
Two-week pilot. Your corpus, your model, your network. We bring the system, deployment support, and the trace IDs your auditor is going to ask about.