The OWASP AI Security Verification Standard (AISVS) provides a structured, community-driven checklist of testable security requirements for AI-driven applications. It targets developers, architects, and security professionals, offering a framework to verify AI system security throughout its lifecycle, from data collection to retirement, ensuring robust and verifiable security controls.

How It Works

AISVS is modeled after the OWASP Application Security Verification Standard (ASVS), offering a catalogue of verifiable, testable, and implementable security requirements for AI systems. It employs three distinct verification levels (1: baseline, 2: standard, 3: high-assurance) to accommodate varying risk profiles and assurance needs. This approach allows organizations to select appropriate security controls based on the sensitivity of data handled and the criticality of the AI application.

Quick Start & Requirements

• Primary access: The latest stable version is AISVS 1.0, available in PDF and Markdown source formats.
• Links: Official documentation and source can be found via the OWASP AISVS project page.
• Prerequisites: None for using the standard itself; it serves as a framework for verifying AI applications.

Highlighted Details

• Verification Levels: Three distinct levels (1, 2, 3) cater to varying assurance requirements, from essential baselines to high-assurance environments.
• Comprehensive Scope: Covers critical AI security domains including data integrity, input validation, model lifecycle management, infrastructure, access control, supply chain, model behavior, memory/vector databases, orchestration, and adversarial robustness.
• Research Wiki: An extensive wiki accompanies the standard, detailing threats mitigated, verification approaches, tooling, and implementation caveats for all 191 requirements.
• Standard Complementarity: Explicitly designed to complement and extend concepts from OWASP ASVS, OWASP Top 10 for LLMs/Agentic Applications, NIST AI RMF, and ISO/IEC 42001.

Maintenance & Community

Project leadership includes Jim Manico, Otto Sulin, Rico Komenda, Russ Memisyazici, and Raza Sharif. The project welcomes community contributions via GitHub issues and pull requests, with a dedicated security policy for reporting vulnerabilities.

Licensing & Compatibility

The project content is licensed under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0). This license permits use, modification, and distribution, provided attribution is given and derivative works are shared under the same license, which may impose copyleft obligations.

Limitations & Caveats

AISVS is explicitly not a governance or risk management framework, nor is it a tool recommendation list. It focuses solely on providing testable technical security requirements. Requirement identifiers may change between versions, necessitating the use of version-tagged identifiers (e.g., v1.0-C9.4.3) for stable referencing.