Skip to main content
AI SECURITY STATIC SCANNER

Detect AI Attack Surface
Before Deployment

Static code analysis for AI/ML security risks

91 Semgrep rules across 78 display IDs and 12 attack categories. Deterministic, reproducible, SARIF-compatible.

91 Semgrep Rules82 Compliance MappingsSARIF Output
91
Semgrep Rules
82
Compliance Mappings
0
AI Guessing
100%
Reproducible
⚠️ Critical Risk

AI Systems Have Unique Attack Surfaces

Traditional security scanners miss AI-specific vulnerabilities like prompt injection, RAG poisoning, and tool abuse.

✗

Traditional Security Scanners

  • Miss AI-specific attack patterns
  • No prompt injection detection
  • No RAG poisoning checks
  • No LLM tool abuse detection
✅

HAIEC AI Static Security Scanner

  • 91 Semgrep rules across 78 display IDs
  • Prompt injection pattern matching
  • RAG poisoning detection
  • Tool abuse and agent safety checks

What We Detect

12 attack categories, 91 Semgrep rules (78 display IDs), 82 compliance mappings across 9 frameworks

Prompt Injection

Critical

User input reaches LLM prompts without validation

Rules:R1.1-R1.7 (7 rules)
Maps to:SOC 2 CC7.2/CC8.1, ISO 27001 A.12.6.1, OWASP LLM01, CWE-77

REST API Detection

Medium

Direct AI REST API calls without SDK wrapper or protection

Rules:R2.1-R2.8 (8 rules)
Maps to:SOC 2 CC7.2/CC8.1, ISO 27001 A.12.6.1, OWASP LLM01

SDK Detection

Info

AI SDK imports detected for system monitoring and change management

Rules:R3.1-R3.5 (5 rules)
Maps to:SOC 2 CC7.2/CC8.1, ISO 27001 A.12.6.1

Dangerous Tool Abuse

Critical

Dangerous tools (PythonREPL, Shell, eval/exec) exposed to AI agents

Rules:R4.1-R4.9 (9 rules)
Maps to:SOC 2 CC6.1/CC7.3, ISO 27001 A.9.2.3, OWASP LLM07, CWE-94

API Key & Secrets Exposure

Critical

Hardcoded API keys, secrets in logs, URLs, or error messages

Rules:R5.1-R5.7 (7 rules)
Maps to:GDPR Art 32, HIPAA 164.312(b), SOC 2 CC6.6, CWE-200

Agent & Advanced Safety

High

Agent loops, recursive calls, missing guardrails, memory injection

Rules:R6.1-R6.10 (10 rules)
Maps to:SOC 2 CC6.1/CC7.2, ISO 27001 A.9.4.1, OWASP LLM09, CWE-674

Data Leakage & Privacy

Critical

PII in prompts, sensitive DB fields, training data leakage

Rules:R7.1-R7.6 (6 rules)
Maps to:GDPR Art 32, HIPAA 164.312(b), SOC 2 CC6.6, CWE-200

RAG & Vector Store Security

High

Unvalidated vector store ops, user-controlled embeddings, metadata injection

Rules:R8.1-R8.6 (6 rules)
Maps to:SOC 2 CC6.1, OWASP LLM03, CWE-20

Production Security & Config

Medium

Missing rate limits, auth, cost tracking, debug mode, error logging

Rules:R9.1-R9.8 (8 rules)
Maps to:SOC 2 CC6.1/CC7.2/CC8.1, ISO 27001 A.9.2.3, CWE-306/CWE-770

Model Security

High

Model extraction, unverified loading, poisoning, weights exposure

Rules:R10.1-R10.5 (5 rules)
Maps to:SOC 2 CC6.1/CC6.6, ISO 27001 A.9.4.1, CWE-494/CWE-200

Injection & XSS via AI Output

High

AI-generated content used unsafely in HTML, SQL, or HTTP requests

Rules:R11.1-R11.3 (3 rules)
Maps to:SOC 2 CC6.1/CC7.2, OWASP A03:2021, CWE-79/CWE-89/CWE-918

Multimodal & Misc

Medium

Image/audio input injection, AI filesystem access, verbose errors

Rules:R12.1-R12.4 (4 rules)
Maps to:SOC 2 CC6.1/CC7.2, OWASP LLM01, CWE-20/CWE-209

Why HAIEC AI Static Security Scanner

Deterministic, reproducible, compliance-ready

Pre-Deployment Detection

Find AI security risks before code reaches production

Compliance Evidence

Automatic mapping to SOC 2, ISO 27001, ISO 42001, GDPR, HIPAA, OWASP, CWE, NIST AI RMF, EU AI Act

CI/CD Integration

GitHub Actions, GitLab CI, Jenkins with SARIF output

Deterministic Results

Same code = same findings, every time. No AI guessing.

System Architecture

How the AI Static Scanner Is Built

Five layers from repo clone to compliance evidence. Source code is ephemeral — always deleted after scan.

Security Blueprint

Repo Handling
Ephemeral clone → delete
Detection
91 rules / 78 display IDs
Compliance
SOC2, ISO, OWASP, CWE
Artifacts
Attestation + Badge + SARIF
Authorization
Extraction
Analysis
Evidence
Output
End-User Process Flow

Your Journey to AI Security Evidence

8 steps from scan initiation to compliance evidence. Source code is ephemeral — never persisted.

STEP 1 OF 8SETUP

Initiate Scan

< 1 min
USER ACTION
What You Do

Provide your GitHub repository URL. Authenticate with your HAIEC account. The scan can be triggered from the dashboard, CI/CD pipeline (GitHub Action), or API.

What the System Does

Scan authorization validates session, checks rate limits, and verifies GitHub access token. Scan intent schema enforced. State machine transitions to authorized.

Output

Authorized scan session with validated repository target

Start Scanning Your AI Code Today

Free scan. No credit card required. Results in minutes.