Skip to main content
ControlFrameby LockedIn Labs
Framework library

Requirements & evidence

OWASP Agentic Top 10

OWASP's peer-reviewed risk framework for autonomous and agentic AI applications, including systems that plan, use tools, hold memory, or coordinate multi-step workflows. It is guidance, not a certification.

Edition on record
2026
Edition published
Source last checked

Understand the scope

What the framework covers.

The library records this framework for AI, technology, SaaS, across global. The issuing authority determines applicability; your assessment scope should identify the organization, systems, services, and edition being assessed.

Areas of requirement

  • Agent goal and context integrity
  • Tool and privilege misuse
  • Agentic supply chain
  • Memory and identity security
  • Human oversight and accountability

10 source entries on record

Risk categories in the OWASP Top 10 for Agentic Applications 2026. These are risk categories, not certification controls, and are not ingested here.

Source access and licensing

OWASP publishes the guide openly. Preserve the publication's attribution and license terms before reproducing or adapting its content in a product catalog.

Prepare with context

The evidence behind
the requirements.

Evidence needs to demonstrate how a requirement is met within the assessment scope. Keep the source requirement, collection context, responsible owner, review decision, and period of applicability together.

A framework-specific collection workflow has not been activated. The reference material on this page can inform scoping and preparation; it does not establish evidence collection or assessment readiness.

ControlFrame availability

Framework reference available

Authority and version facts, with a parsed public catalog when available; no tenant blueprint or executable evidence method is activated.

How availability is determined

OWASP Agentic Top 10 is tracked in the registry — authority, version ledger, verification — and nothing else is modelled for it yet. Import its catalog to begin.

Catalog status: Planned. Tracked metadata: we record the regime and its issuing authority for review. Registry presence does not mean automated monitoring is active. No control or requirement model is implemented.

Metadata only; risk entries, threat scenarios, and mitigations are not ingested or released as an executable agent-security module.

Follow the source

Editions and transitions.

These are the editions recorded in the library, checked August 30, 2026. Confirm the applicable edition with the authority before beginning an assessment.

  • 2026Current edition on record

Check the basis

Sources you can follow.

Verified against the issuing authority’s own publication. Source verification was recorded on August 30, 2026.

Catalog provenance and record details

Framework record: owasp-agentic-top-10-2026.

Source artifact: acquisition required.

Parsed catalog: Not ingested.

Control families as published: Agent goal and context integrity · Tool and privilege misuse · Agentic supply chain · Memory and identity security · Human oversight and accountability.

Record last updated: August 30, 2026.

Common questions

A closer look.

What is OWASP Agentic Top 10?

OWASP's peer-reviewed risk framework for autonomous and agentic AI applications, including systems that plan, use tools, hold memory, or coordinate multi-step workflows. It is guidance, not a certification.

Who does OWASP Agentic Top 10 apply to?

OWASP Agentic Top 10 applies to AI, technology, SaaS, global, per OWASP GenAI Security Project.

What is the current version of OWASP Agentic Top 10?

The current edition is 2026, issued by OWASP GenAI Security Project and published 2025-12-09. Source: https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/.

What does an assessment under OWASP Agentic Top 10 require?

10 control units are on record (Risk categories in the OWASP Top 10 for Agentic Applications 2026. These are risk categories, not certification controls, and are not ingested here.), organized into 5 control families: Agent goal and context integrity, Tool and privilege misuse, Agentic supply chain, Memory and identity security, Human oversight and accountability.

At a glance

OWASP Agentic Top 10

2026

Framework reference
Authority and edition tracked
Evidence preparation
Framework reference available
Source confidence
Primary authority

A framework reference describes requirements. Evidence decisions belong to a scoped assessment and a named reviewer.

Continue exploring

Related frameworks.

CSA AI Controls Matrix

Also applies to AI · The Cloud Security Alliance's vendor-neutral control framework for cloud-based AI systems, with implementation and auditing guidance plus the companion AI-CAIQ. It is a control catalog, not by itself a certification or STAR for AI designation.

AIUC-1

Also applies to AI · A quarterly updated standard and certification program for AI agents covering data and privacy, security, safety, reliability, accountability, and societal risk. Only AIUC can issue its certificate; registry inclusion makes no certification claim.

OWASP GenAI LLM Top 10

Also applies to AI · OWASP's current community-driven guide to the most critical security risks for applications powered by large language models. It is security guidance, not a compliance certification or organizational assurance report.

CSA STAR

Also applies to SaaS · The Cloud Security Alliance's cloud-assurance program and public registry. Level 1 is a CCM/CAIQ self-assessment; Level 2 is a third-party certification or attestation path. Tracking STAR does not place ControlFrame or any customer on the registry.

Make the requirements
part of the work.

Explore how ControlFrame can organize OWASP Agentic Top 10 requirements, evidence preparation, and review for your assessment scope.

Request a walkthrough
OWASP Agentic Top 10 | ControlFrame