AI SAFE² v3.0 — Interactive Framework Dashboard

161

Total Controls

Across 5 pillars + governance

32

Frameworks Mapped

ISO, NIST, EU AI Act + 29 more

3

First-in-Field Controls

CP.7, CP.9, CP.10 — no other framework has these

4

ACT Capability Tiers

Governance scales with agent autonomy

Framework Architecture

6 governance domains

P1 Sanitize

P2 Audit

P3 Fail-Safe

P4 Monitor

P5 Evolve

CP Governance

What Each Domain Protects

CP.1–CP.10: Cross-Pillar Governance OS

NEW IN v3.0

Ten governance controls that operate across all five pillars simultaneously — not within any single domain. Includes three first-in-field standards that no other AI governance framework has published.

Compliance Coverage

32 frameworks in one implementation

AI-Specific

Enterprise

Regulatory

+14 more

CSA, MAESTRO, Google SAIF, MIT AI Risk, AIID, CCPA/CPRA, NIST CSF 2.0 and 7 more

Board-level point: One AI SAFE² v3.0 implementation generates evidence artifacts for 32 compliance frameworks simultaneously. Your team audits once, satisfies everything.

Ready to present your AI risk posture?

The AI SAFE² Implementation Toolkit includes the Risk Command Center dashboard, 161-point audit scorecard, HEAR designation forms, and board-ready compliance evidence packages.

Get the Toolkit — $97 ☆ Star on GitHub

ACT Capability Tier Classifier

CP.3 v3.0

Answer 6 questions about your agent. Receive your ACT tier, mandatory controls, and governance requirements.

⚠

CP.10 HEAR Designation Required

This tier requires a named Human Ethical Agent of Record with a cryptographic signing key before production deployment. This is a first-in-field governance standard — no other framework defines this.

⚠

CP.9 Agent Replication Governance Required

Orchestrators spawning sub-agents must implement lineage tokens, delegation hop limits (max 3), ephemeral credentials, and 500ms kill-switch tree. First-in-field: no other framework covers agent replication.

🛑

CP.8 Catastrophic Risk Thresholds — Deployment Blocker

Document behavioral indicators that trigger emergency suspension before deployment is approved. Required for all ACT-3 and ACT-4 agents.

Mandatory Controls for Your Tier

Next step: Generate your complete deployment governance package including HEAR designation form, CP.9 lineage spec, and audit evidence checklist. Get the Toolkit →

✅

Free: AI Builder Pre-Flight Checklist

35 questions — before you ship any AI agent to production

FREE

Your ACT tier tells you what governance you need. The Pre-Flight Checklist tells you whether your agent is actually ready to deploy. 35 structured questions across every critical risk surface — security, data governance, human oversight, failure modes, and compliance readiness.

Get the Checklist — Free → No account required · GitHub download · 35 questions · Instant access

Builder Reference — Controls by Impact

🚀 FREE RESOURCE

CSI AI Builder Pre-Flight Checklist

35 structured questions that answer one question before you ship: is this agent actually ready? Covers security, governance, human oversight, failure modes, and compliance. Built directly from AI SAFE² v3.0 controls — the same framework you are using right now.

Download Free Checklist

✓ No email required on GitHub

✓ 35 questions · PDF + Markdown

✓ Maps to AI SAFE² v3.0 controls

Sample questions

+ 31 more questions

Scan your codebase now

The AI SAFE² v3.0 scanner checks your code against 40+ rules covering all 5 pillars and CP.1-CP.10. ACT tier estimated from your code structure.

python -m scanner.cli scan . --tier Tier2 --report json

View on GitHub →

Compliance Framework Crosswalk

⚖

Select a compliance framework above

See which AI SAFE² v3.0 controls satisfy each requirement

controls

Audit tip: Export the complete gap report with A2.5 execution trace requirements, compliance evidence artifact types, and control implementation status. Export report →

Self-Assessment Scorecard

Check controls you have implemented. Score updates live.

CRITICAL Controls

HIGH Controls

Export your results: The full 161-point audit scorecard with auto-calculated Combined Risk Score, ISO 42001 alignment, and compliance evidence package is in the Toolkit. Get Toolkit →

v3.0 Combined Risk Score Calculator

Score = CVSS_Base + ((100 − Pillar_Score) ÷ 10) + (AAF ÷ 10)

CVSS Base Score

0 10

Vulnerability severity (CVE base score)

Pillar Score (AI SAFE² compliance)

0 100

From your assessment above (100 = fully compliant)

OWASP AIVSS AAF (0-100)

0 100

Agentic Amplification Factor (10 factors × 0-10)

CVSS and AAF are estimates from static analysis. Use the MCP server risk_score tool for precise calculation.

CROSS-PILLAR GOVERNANCE OS ALL NEW IN v3.0

CP.1 – CP.10: First-in-Field Standards

Ten governance controls that operate across all five pillars simultaneously. Three have no counterpart in any existing AI governance framework.

Attack Surface Map — Controls by Threat Category

MITRE ATLAS Crosswalk

23 New v3.0 Controls

ADDED IN v3.0

Who are you here as?