Correctover · AI Reliability™ — Agent Runtime Assurance

AISVS 1.0 Coverage Map

OWASP AISVS v1.0 (released 24 June 2026) defines 12 requirement chapters across the AI lifecycle, with requirements at verification levels 1–3. The free scanner runs 31 static checks (14 for MCP configurations, 17 for code/bundles). This page states honestly which chapters those checks touch, and which they do not.

References: OWASP/AISVS repository · official AISVS documentation. Chapter names below follow the official v1.0 text.

Chapter coverage at a glance

ChapterAISVS v1.0 requirement chapterFree scanner
C1Training Data Integrity & Traceability○ Not covered
C2Input Validation (prompt-injection defenses, content screening)● Static signals
C3Model Lifecycle Management & Change Control○ Not covered
C4Infrastructure, Configuration & Deployment Security● Static signals
C5Access Control & Identity● Static signals
C6Supply Chain Security for Models, Frameworks & Data◐ Partial signals
C7Model Behavior, Output Control & Safety Assurance◐ Partial signals
C8Memory, Embeddings & Vector Database Security○ Not covered
C9Autonomous Orchestration & Agentic Action Security● Static signals
C10Model Context Protocol (MCP) Security● Static signals
C11Adversarial Robustness & Attack Resistance○ Not covered
C12Monitoring, Logging & Anomaly Detection◐ Partial signals

What the 31 checks actually look for

ChapterChecks (static patterns, with file/line locations)
C10 MCP Security TLS / plaintext HTTP endpoints MCP transport authentication SSRF: cloud metadata & internal addresses (169.254.169.254, RFC1918 ranges)
C9 Agentic Action Timeout / interruption signals Tool allowlists & permission gates Kill switch / AbortController Token / cost budgets (semi-automated)
C5 Access Control Hardcoded credentials & API keys Credential flow via environment variables
C4 Infrastructure Subprocess execution (shell:true, exec) Dynamic code execution (eval / Function / vm) Sandbox / isolation mechanisms
C2 Input Validation Input validation & schema enforcement
C7 Output Control Output truncation / filtering (semi-automated)
C6 Supply Chain Dependency version pinning / vendor SBOM signals (semi-automated)
C12 Monitoring Audit logging / telemetry (semi-automated) Error handling & retry policy (semi-automated)
● automated pattern detection · ◐ semi-automated flags candidate code regions for human confirmation (library code legitimately contains such patterns; context determines whether a gate exists).

What the free scanner deliberately does not claim

C1 (training data), C3 (model lifecycle/change control), C8 (memory, embeddings, vector databases) and C11 (adversarial robustness) cannot be assessed by static pattern matching on client code and MCP configs — they concern data provenance, model artifacts, embedding pipelines and runtime attack resistance. Even within the 8 chapters touched, the scanner sees surface signals only: it does not prove reachability, exploitability, or whether a flagged pattern is a real vulnerability versus a safe/legitimate use. Library code (including this scanner itself) contains patterns like child_process and credential-matching regular expressions that are detection logic, not vulnerabilities — expect false positives, and false negatives.

The 116-check manual deep audit works from the scanner's raw signals and traces each one through real call paths — reachability, exploitability, semantic intent, prioritized remediation — under a signed engagement. It still does not certify AISVS; it delivers an opinion scoped to the code and timeframe agreed.

⚠ Legal notice

Not an audit or certification. The free scanner's output is automated static analysis. It is not a security audit, penetration test, certification, or statement of compliance with AISVS or any other standard. No score — including a clean scan — constitutes evidence that a system is secure, vulnerability-free, or compliant.

OWASP trademarks. OWASP® and AISVS are trademarks of the OWASP Foundation. Correctover is an independent project, not affiliated with or endorsed by OWASP. OWASP does not certify products or vendors. The chapter mappings on this page are Correctover's own, provided for navigation only.

Use at your own risk. The scanner and reports are provided "as is", without warranty of any kind. To the maximum extent permitted by law, Correctover shall not be liable for damages arising from their use. Do not rely on a free scan as the sole basis for acceptance, procurement, release, insurance, or regulatory decisions.

▶ Run the free in-browser scanner Learn about the 5-day manual deep audit