Skip to main content
SaaSVista
v1.0 — Assessed 2026-03-23

We practice what we preach.

The first AI governance company to publish its own assessment. Here's exactly how we handle your code, what we send to AI models, and what we never will.

Self-Assessment Summary

527

AI Detections

33

Findings

48

Controls Mapped

2,230,466

Lines Scanned

Critical: 1 High: 12 Medium: 20
NIST AI RMFISO 42001EU AI Act anthropicopenaigoogle_aigroqlangchainlanggraphaws_bedrockhuggingfacemcpsemantic_kerneltogether_aideepseek

Three Tiers of Data Privacy

SaaSVista Scan

Standard

Cloud-hosted scan — your code stays on GitHub

  • ✓ Read-only GitHub API access (fine-grained PAT)
  • ✓ Code analyzed in isolated containers
  • ✓ Findings stored encrypted at rest (LUKS2 AES-256)
  • ✓ No code persisted after scan — only findings retained

LLM: Claude API (Anthropic) for enrichment

SaaSVista Private

Private

On-prem VM appliance — your code never leaves your network

  • ✓ Runs on your infrastructure (VMware/KVM)
  • ✓ Air-gap capable — no internet required
  • ✓ Embedded AI model (no external API calls)
  • ✓ You control all data, all the time

LLM: Embedded Qwen2.5-7B (runs locally, no cloud)

SaaSVista Sovereign

Sovereign

Full data sovereignty — code snippets redacted from reports

  • ✓ Everything in Private tier, plus:
  • ✓ Code snippets auto-redacted from all outputs
  • ✓ Only file:line references in reports
  • ✓ Transparency log proves what was sent to any LLM

LLM: Embedded model + snippet redaction

What We Never Send to Any LLM

✗ API keys, tokens, or secrets
✗ Database connection strings
✗ Customer data or PII
✗ Full source code files
✗ Environment variables
✗ Private key material

Enforced via SecretsFilter (12 regex patterns) applied before any LLM call. Verified by transparency log with SHA-256 hashes.

Response Kit Preview

Sample questions from our 27-question due diligence response package

How does SaaSVista handle customer source code?

Scan Evidenced

Read-only access via GitHub fine-grained PAT. Code is analyzed in memory, never persisted. Only structured findings (file paths, detection types, severity) are stored.

What AI models does SaaSVista use?

Scan Evidenced

Claude (Anthropic) for finding enrichment in hosted mode. Embedded Qwen2.5-7B for on-prem appliance. All LLM calls are logged in a transparency log with SHA-256 hashes.

Does SaaSVista comply with NIST AI RMF?

Scan Evidenced

Yes — 23/23 NIST AI RMF controls are mapped. Our scan produces evidence against GOVERN, MAP, MEASURE, and MANAGE function areas.

Can SaaSVista run without internet access?

Confirmed

Yes — the Private and Sovereign tiers run on-prem with an embedded AI model. No external API calls required. Air-gap deployment supported.

Want to see what your governance posture looks like?

Book a Governance Assessment