AI AppSec v0.1.0

AI AppSec v0.1.0 — Release Notes

What AI AppSec Is

AI AppSec is an evidence-backed application security tool for AI applications
and agents. It performs deterministic static analysis on AI-agent and
AI-application source code, producing reproducible security findings with
coverage receipts and evidence envelopes.

Why Evidence-Backed AppSec

AI applications and agents introduce new security surfaces: prompt injection,
tool misuse, data exfiltration, insecure model integrations, and agent
orchestration risks. AI AppSec provides deterministic, reproducible evidence
so developers and security engineers can audit AI code before commit, push,
merge, or deployment — without overclaiming coverage or certainty.

What v0.1.0 Includes

• scan_ai_security — MCP tool for AI source-code security auditing
• 122 Public Core detectors — Semgrep-based static analysis rules
• 79 semantic security checks — higher-level security concern detection
• Security Concern Families — organized finding classification
• COMPLETE / PARTIAL / ERROR — explicit coverage semantics
• Scan Receipts — SHA-256 digests for deterministic reproducibility
• Evidence Envelopes — structured evidence output
• Proof-of-fix — rescan comparison with check-evaluation safety
• Local/offline scanning — no network required after Semgrep setup
• MCP stdio transport — integrates with MCP-compatible AI tools
• Node 22/24 support — Node 26 canary qualified

Installation

npm install -g ai-appsec

# Check Semgrep status
ai-appsec doctor

# Install Semgrep engine (requires network)
ai-appsec setup

Or use via npx:

npx ai-appsec doctor
npx ai-appsec setup

MCP Configuration

Add to your MCP client configuration:

{
  "mcpServers": {
    "ai-appsec": {
      "command": "npx",
      "args": ["ai-appsec"]
    }
  }
}

MCP identity: io.github.subodhkc/ai-appsec

Example Natural-Language Requests

• "Check this code for security issues"
• "Audit this AI agent before I push"
• "Review this change for security problems"
• "Scan this AI application"
• "Is this safe to merge?"
• "Audit this code for vulnerabilities"

Evidence Semantics

• COMPLETE: All supported files in scope were successfully analyzed
• PARTIAL: Some files could not be analyzed (reported explicitly)
• ERROR: The scan could not complete due to an engine error

Missing evidence is reported as missing — never as success.

Known Limitations

• Static analysis only — does not execute target code
• Does not provide complete AI system assurance
• Does not provide runtime security monitoring
• Does not provide compliance certification
• Semgrep engine must be installed separately (ai-appsec setup)
• Network required only for Semgrep engine setup, not for normal scans

Security / Reporting

• Deterministic findings — same input produces same output
• Fail-closed digests — rulepack/manifest verified at runtime
• No HAIEC account or API key required
• No telemetry or phone-home behavior
• Scan Receipts enable reproducible evidence sharing

License

MIT — Copyright (c) 2026 HAIEC

Roadmap

• scan_tenant_isolation — cross-tenant boundary checks
• verify_llm_content — LLM input/output verification
• check_deploy_security — release/deployment orchestration gate
• MCP Registry stable entry
• Platform integration (future phase)