Knowledge Bases and Evaluation Organizations: OWASP GenAI Security Project
Knowledge Bases and Evaluation Organizations · OWASP GenAI Security Project. LLM Top 10, Agentic Top 10 and ACS standards. This source has 8 entries admitted under this category's review policy; each links directly to the original.
Authority tiers in this category: A2–B. Admission policy: All core resources are included. Each secondary source is admitted according to its own database, methodology, versions and leaderboard structure; there is no cross-organization scoring scheme.
OWASP GenAI LLM Top 10 2026
Ten core security risks for LLM applications with mitigation advice; the 2026 edition updates the risk ranking with real incident data.
Evidence use: Supports classification of LLM application risks, threat descriptions and mitigations.
Limitations: A community guide, not a mandatory standard. Must be pinned to the 2026 version.
Related concepts: AI Guardrails · AI Red Teaming
Link last checked: 2026-09-24
OWASP Top 10 for Agentic Applications 2026
Summarizes key risks for autonomous agents, including goal hijacking, tool misuse, identity and permissions, supply chain and code execution.
Evidence use: Supports threat modeling and security control design for agent applications.
Limitations: Focuses on application-layer risks. Other agent architectures require their own contextual analysis.
Related concepts: AI Agent · AI Guardrails · AI Red Teaming
Link last checked: 2026-09-24
Agent Control Standard（ACS）
Establishes an open control foundation for observability, middleware control hooks and runtime policy enforcement on agent platforms.
Evidence use: Supports designing runtime controls, auditing and portable security policies for agents.
Limitations: A newly released standard; its implementation ecosystem is still forming. Cannot replace the permission model of a specific system.
Related concepts: AI Agent · AI Guardrails · AI Governance and Regulation
Link last checked: 2026-09-24
OWASP GenAI Red Teaming Guide
Sets out red teaming scope, test phases, attack techniques and reporting practices for generative AI systems.
Evidence use: Supports designing reusable adversarial testing processes for generative AI.
Limitations: An open community guide, not a mandatory regulation. Adoption must pin a version and account for the specific architecture.
Related concepts: AI Red Teaming
Link last checked: 2026-09-24
OWASP Multi-Agentic System Threat Modeling Guide v1.0
Maps trust boundaries, communication, permissions and cascading failure risks in multi-agent systems.
Evidence use: Supports systematic threat modeling of multi-agent architectures.
Limitations: An open community guide, not a mandatory regulation. Adoption must pin a version and account for the specific architecture.
Related concepts: AI Agent · AI Red Teaming
Link last checked: 2026-09-24
OWASP GenAI Incident Response Guide 1.0
Covers preparation, detection, containment, recovery and post-incident review for generative AI incidents.
Evidence use: Supports establishing AI-specific incident response processes and evidence preservation requirements.
Limitations: An open community guide, not a mandatory regulation. Adoption must pin a version and account for the specific architecture.
Related concepts: AI Governance and Regulation · AI Guardrails
Link last checked: 2026-09-24
OWASP GenAI Security Industry Framework Crosswalk
Maps OWASP GenAI risks to the control requirements of major security and governance frameworks.
Evidence use: Supports reusing controls across frameworks and gap analysis.
Limitations: An open community guide, not a mandatory regulation. Adoption must pin a version and account for the specific architecture.
Related concepts: AI Governance and Regulation
Link last checked: 2026-09-24
OWASP AI Security Solutions Landscape
Organizes AI security products and open-source solutions by protective capability and application stage.
Evidence use: Supports identifying categories of AI security controls and tool coverage.
Limitations: An open community guide, not a mandatory regulation. Adoption must pin a version and account for the specific architecture.
Related concepts: AI Guardrails
Link last checked: 2026-09-24