AI-SHIELD — AI Security & Trust Maturity Assessment

Independent project

The challenge

Every major AI security framework published in the last two years — NIST’s AI RMF, ISO/IEC 42001, the Cloud Security Alliance’s 243-control AI Controls Matrix — is built for an organisation with a dedicated security team and months to run a formal audit. Most teams actually shipping AI features don’t have that team or that time, and self-assessments without any structural resistance to optimism tend to score everything a comfortable 4 out of 5.

What I built

Designed and built a free 18-question maturity assessment scoped differently on purpose: eighteen questions a person who actually built or operates the system can answer honestly in fifteen minutes, across six pillars — Supply Chain & Vendor Trust, Human Oversight & Governance, Identity, Access & Agency, Evaluation & Robustness, Logging, Detection & Response, and Data & Privacy. Every answer level describes a concrete, checkable state rather than a feeling, a governance cap stops one strong pillar from masking a critically weak one, and a single zero anywhere is flagged as a named Critical Gap regardless of the overall average. Results are grounded throughout with citations into NIST, ISO, OWASP, and CSA rather than presented as a standalone opinion.

Outcomes

• 18 evidence-anchored questions across six pillars, each mapped to specific citations in NIST’s AI RMF, ISO/IEC 42001, the OWASP LLM/ASI Top 10, and the Cloud Security Alliance’s AI Controls Matrix
• A governance cap that resists self-assessment inflation — the overall score can’t exceed Human Oversight & Governance maturity by more than one point
• Exports a board/audit-ready PDF with ranked gaps, owners, and target dates, or a one-page executive summary
• Supports multiple named, scoped assessments per organisation with a portfolio view, re-assessment delta tracking, and an opt-in anonymous benchmark against other respondents

Tech & approach

Claude Code · Static HTML/CSS/JS · No backend, no dependencies